A method and device for optimizing double-layer cost decision of a thermal power plant

Abstract

The application discloses a kind of thermal power plant double-layer cost decision optimization method and device, belong to thermal power plant cost optimization field.Therein method includes: establishing auxiliary database;According to auxiliary database design minimum coal purchase / stock cost function, calculate the lowest standard unit price of entering furnace, construct double-layer cost decision optimization model;According to the characteristics of double-layer cost decision optimization model, by McCormick-KKT is converted into single-layer model, by MATLAB is solved, obtains the optimal cost of base model;According to the optimal cost of base model, determine the expected cost, build IGDT model into different risk attitudes, to obtain risk decision optimization and robustness decision optimization;According to risk decision optimization and robustness decision optimization, calculate risk robustness-cost curve.The application overcomes the problems existing in the prior art, such as the cost optimization technology being difficult to perceive future coal spot market fluctuations, being mostly local static optimization, and having a narrow decision-making perspective.

Description

Technical Field

[0001] This invention relates to the field of cost optimization in thermal power plants, and more particularly to a two-level cost decision optimization method and apparatus for thermal power plants. Background Technology

[0002] In recent years, under the dual-carbon context, low-marginal-cost renewable energy sources such as wind and solar power have entered the electricity market, while the prices of bulk energy sources such as coal and natural gas have fluctuated dramatically. Thermal power plants, with fixed power generation quotas, experience relatively small fluctuations in electricity sales revenue, while the sharp rise in coal costs has led to large-scale losses and low willingness to generate electricity. There is an urgent need to adjust current production decisions to tap potential for reducing power generation costs and ensuring the safe supply of electricity to the power system.

[0003] Most existing technologies neglect feedback on coal purchasing decisions. They only consider off-site coal purchases and storage or only focus on blending ratios for cost optimization, lacking a holistic approach that integrates both factors and ignores the potential for production reduction in thermal power plants through coordinated optimization. Spot coal price forecasting, however, offers decision-makers more possibilities. Given the significant uncertainty surrounding future coal prices due to supply constraints, energy market fluctuations, and policy documents, there is a pressing need to reasonably measure the decision-making risks associated with forecasting coal prices to improve decision robustness and risk management capabilities.

[0004] Therefore, from the perspective of short- to medium-term coal price forecasting, measuring market uncertainty risk decisions based on information gap decision-making theory and achieving dual-layer dynamic linkage decision optimization of coal purchase inventory and blended coal combustion can help power generation companies reduce costs and increase profits in the context of large fluctuations in coal prices. Summary of the Invention

[0005] In order to at least partially solve one of the technical problems existing in the prior art, the purpose of this invention is to provide a two-level cost decision optimization method and apparatus for thermal power plants.

[0006] The technical solution adopted in this invention is:

[0007] A two-level cost decision optimization method for thermal power plants.

[0008] Includes the following steps:

[0009] S1. Establish an auxiliary database; wherein the auxiliary database includes coal consumption characteristic curves of thermal power units and spot coal market information data.

[0010] S2. Based on the auxiliary database, design the minimum coal purchase / inventory cost function, calculate the minimum standard unit price for furnace input, and construct a two-level cost decision optimization model that measures the linkage between coal purchase and inventory in the spot market.

[0011] S3. Based on the characteristics of the two-level cost decision optimization model, it is transformed into a single-level model through McCormick-KKT, and solved using MATLAB to obtain the optimal cost of the base model.

[0012] S4. Determine the expected cost based on the optimal cost of the base model, and construct an IGDT model to incorporate different risk attitudes in order to obtain risk tolerance decision optimization and robust decision optimization.

[0013] S5. Based on risk decision optimization and robustness decision optimization, calculate the risk robustness-cost curve, and further revise the coal purchase / blending decision scheme according to actual preferences.

[0014] Furthermore, the establishment of the auxiliary database includes:

[0015] Historical operating data of thermal power units were obtained, and the coal consumption characteristic curves of each unit were obtained using curve fitting methods;

[0016] Based on relevant information about spot coal prices, obtain spot coal market information data; train a spot coal price prediction model based on the spot coal market information data, and use the spot coal price prediction model to predict the trend of spot coal prices;

[0017] Historical power generation information of thermal power units and future spot coal price market information are coupled to form an auxiliary database.

[0018] Furthermore, the historical operating data includes the maximum output, minimum output, start-up cost, shutdown cost, shortest start-up time, shortest shutdown time, maximum ramp rate, and historical standard coal consumption for power generation of the thermal power unit.

[0019] The standard coal consumption curve of the thermal power unit is fitted using the least squares method, where the standard coal consumption curve function satisfies the following equation:

[0020] C t =at 2 +bt+c

[0021] In the formula, C t Let t represent the standard coal consumption of the thermal power unit during the period t. The ratios a, b, and c represent the quadratic, linear, and constant terms of the standard coal consumption curve, respectively.

[0022] The expression for the least squares method is as follows:

[0023]

[0024] In the formula, y t Let f(t,a,b,c) be the historical standard coal consumption for power generation of thermal power units, and let f(t,a,b,c) be the standard coal consumption curve function.

[0025] Furthermore, the step of obtaining spot coal market information data based on spot coal price information includes:

[0026] Based on relevant information about spot coal prices, spot coal market information data is obtained through multiple spot coal price forecasting steps;

[0027] The spot coal price forecasting steps include:

[0028] Data processing: Cleaning and filling in the collected spot coal price-related information data;

[0029] Model building: Using the processed data, a spot coal price prediction model is built. Historical data related to spot coal prices are input for model training to obtain network weights and training parameters.

[0030] Coal price forecast: Using a trained forecasting model combined with relevant information on spot coal prices, the system forecasts future spot coal prices and saves the forecast results.

[0031] Furthermore, the spot coal price prediction model is a coal price prediction model based on the Stacking ensemble learning framework. By integrating multiple intelligent prediction models through the ensemble learning framework, collective intelligence is achieved to improve prediction accuracy.

[0032] The ensemble learning framework includes a primary learner and a meta-learner, and uses K-flod cross-validation to achieve hierarchical coupling.

[0033] Selection of primary learners: The number of dissimilar points and the mean square error in the predictions of each learner are used as measures to select primary learners with different differences and combine them into a model with high generalization performance.

[0034] Determine the meta-learner: The meta-learner must make full use of the differences in predictions among the primary learners. Using the XGBoost model as the meta-learner, K-flod cross-validation is introduced. The cross-validation calculation results of the first-layer primary learner are used as the features of the training set and prediction set of the second-layer meta-learner.

[0035] Furthermore, the construction of a two-tiered cost decision optimization model that links coal purchase and inventory in the spot market includes:

[0036] Selecting coal suppliers by predicting coal prices, and choosing key indicators. The regulations are as follows:

[0037]

[0038] In the formula, and These are priority flags for purchasing spot coal and long-term contract coal, respectively. A value of 1 indicates that the m-period prioritizes the purchase of spot coal. A value of 0 indicates priority for the decomposition of long-term contract coal; λ m,j and These represent the predicted spot coal price for type j over period m and the average long-term contract coal price over a given period.

[0039] The upper-level model optimizes the procurement / inventory management decision based on the best choice between spot coal and long-term contract coal prices, with the total power generation cost as the objective, and transmits the coal purchase plan to the lower-level coal blending and combustion decision optimization; the total cost Y of the power generation enterprise's coal purchase decision is expressed as:

[0040]

[0041] In the formula, Let be the breakdown quantity (decision variable) of the j-th long-term coal contract in the m-th period. Let be the spot coal purchase volume in the m-th period (decision variable); Let j be the spot coal price in the m-th period. The price of the j-th long-term contract coal in the m-th cycle; S is the average warehouse management cost. m,i θ represents the inventory level for this period, θ represents the unit inventory cost, T represents the total number of periods, M represents the number of coal suppliers, and N represents the number of supplier categories.

[0042] The lower-level model optimizes the coal blending scheme in the production process, calculates the unit cost of coal fed into the furnace, and transmits the coal consumption of each cycle to the upper-level coal procurement timing decision optimization; the coal blending ratio optimization model uses the lowest unit blending cost in each cycle. The objective function is:

[0043]

[0044] In the formula, it is assumed that ε m,i Let be the blending ratio of type i coal in the m-th cycle, then the unit blending cost in the m-th cycle is... The unit cost of fuel blending for power generation companies is determined by i; i represents the type of coal, B m,i Let p be the purchase quantity of the i-th type of coal in cycle m, and p be the coal delivery order. m,i , θ represents the average inventory management cost, and θ represents the unit inventory cost. The standard coal calorific value (7000 kcal), Q net,i The calorific value is that of ordinary coal;

[0045] The upper-level optimization aims to minimize coal purchase and inventory costs and then passes the coal purchase plan down to the lower level; the lower-level optimization uses the furnace feed list as the optimization objective to solve for the blending ratio and then passes the consumption of each type of coal up to the upper level.

[0046] Furthermore, the features of the two-level cost decision optimization model include:

[0047] The upper-level model is a linear model, and its classification inventory constraints are constrained by the optimization results of the lower-level blending ratio optimization model.

[0048] The objective function of the lower-level model contains bilinear terms, the form of which is indirectly determined by the solution results of the upper-level model.

[0049] The two-layer model influences each other, and the constructed two-layer optimization model belongs to the two-layer nonlinear optimization model.

[0050] Further, step S3 includes:

[0051] S31. For bilinear terms in the lower-level model, replace the bilinear terms that cause nonlinearity with new variables;

[0052] S32. Add new constraints resulting from variable substitution;

[0053] S33. By combining the old and new constraints, the nonlinear terms containing bilinearity in the original lower-level optimization are transformed into linear terms.

[0054] S34. Use KKT conditions to solve for the extrema of the lower-level model, transforming the lower-level model into a series of constraints, thereby transforming the two-level nonlinear programming into a single-level linear programming.

[0055] S35. Solve a single-level integer linear programming model using MATLAB's CPLEX function.

[0056] Further, step S32 includes:

[0057] According to the McCormick method, linearizing bilinear terms using variable substitution requires the introduction of the following new constraints:

[0058]

[0059] In the formula, x and y are the decision variables being replaced, and w is the decision variable replacing the bilinear term of xy; w m,i This represents the amount of coal of type i decomposed in week m. This represents the minimum blending ratio for coal type i. This represents the maximum blending ratio of coal type i. The minimum purchase quantity of type i coal for week m. This represents the maximum amount of coal purchased for class i during week m.

[0060] Furthermore, the IGDT model constructed in step S4 includes:

[0061] The robust optimization-based IGDT model is represented as follows:

[0062]

[0063] In the formula, The robustness function (which becomes the opportunity-risk function when it reaches its minimum value) has an uncertainty factor α∈[0,1], determined by the coal purchase and blending decision scheme q(W,D,ε) and the expected cost threshold r. c The decision is made; R is the reward function, which is set here as the reciprocal of the cost of the thermal power plant, 1 / Y. The reward function is subject to the initial estimate of uncertainties. In this invention, the factor is the predicted value of spot coal price; U is defined as the set of all uncertain estimates u.

[0064] Using a fractional error model to create an extended interval for measuring future coal price volatility:

[0065]

[0066] In the formula, w is the conversion factor.

[0067] Another technical solution adopted in this invention is:

[0068] A two-level cost decision optimization device for thermal power plants, comprising:

[0069] At least one processor;

[0070] At least one memory for storing at least one program;

[0071] When the at least one program is executed by the at least one processor, the at least one processor performs the method as described above.

[0072] Another technical solution adopted in this invention is:

[0073] A computer-readable storage medium storing a processor-executable program, which, when executed by a processor, performs the method described above.

[0074] The beneficial effects of this invention are: it overcomes the problems of existing cost optimization technologies, such as difficulty in perceiving future fluctuations in the coal spot market, mostly being local static optimizations, and having a narrow decision-making perspective; in addition, it provides decision-makers with more flexible decision optimization schemes by conducting decision risk assessment based on IGDT. Attached Figure Description

[0075] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0076] Figure 1 This is a flowchart of a two-level cost decision optimization method for thermal power plants based on IGDT for assessing decision risk in an embodiment of the present invention;

[0077] Figure 2 This is a schematic diagram of the standard coal consumption characteristics of the generator set in an embodiment of the present invention;

[0078] Figure 3 This is a schematic diagram of the spot coal and long-term contract coal delivery orders for April 2021 in an embodiment of the present invention;

[0079] Figure 4 This is a schematic diagram of a short-to-medium term thermal coal price forecasting model based on Stacking-EL in an embodiment of the present invention;

[0080] Figure 5 This is a flowchart of the conversion method for a two-layer model using the McCormick-KKT method in an embodiment of the present invention;

[0081] Figure 6 This is a comparison chart of total inventory and category inventory in different scenarios in embodiments of the present invention;

[0082] Figure 7 This is a comparison chart of cost indicators for scenarios 3 and 4 in this embodiment of the invention;

[0083] Figure 8 This is a comparison chart of the coal purchasing decision results in scenarios 4 and 5 of this invention.

[0084] Figure 9 This is a risk robustness-cost curve calculated using the IGDT model in an embodiment of the present invention. Detailed Implementation

[0085] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0086] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0087] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0088] Furthermore, in the description of this invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0089] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0090] Terminology Explanation:

[0091] McCormick-KKT: Karush-Kuhn-Tucker conditions based on the McCormick envelope.

[0092] CPLEX: A commercial solver.

[0093] IGDT: Info-Gap Decision Theory.

[0094] like Figure 1 As shown, this embodiment provides a two-layer cost decision optimization method for thermal power plants based on IGDT for assessing decision-making risks. This method overcomes the problems of existing thermal power plant cost optimization technologies, such as difficulty in perceiving future coal spot market fluctuations, primarily involving localized static optimization, and narrow decision-making perspectives. The method specifically includes the following steps:

[0095] S1. Establish an auxiliary database.

[0096] Historical operating data of coal-fired power plants were obtained, and the standard coal consumption characteristic curves of each unit were obtained using curve fitting methods. Based on spot coal price information, spot coal market information data were obtained through multiple spot coal price prediction steps.

[0097] As an optional implementation method, historical operating data of thermal power units in power generation enterprises includes maximum output, minimum output, start-up cost, shutdown cost, shortest start-up time, shortest shutdown time, maximum ramp rate, and historical standard coal consumption for power generation. Based on historical operating data, the standard coal consumption curve of the unit is fitted using the least squares method. A typical standard coal consumption curve function satisfies the following equation:

[0098] C t =at 2 +bt+c

[0099] In the formula, C t Let t represent the standard coal consumption of the generating unit during time period t. Let a, b, and c represent the quadratic, linear, and constant coefficients of the standard coal consumption curve, respectively, obtained by least squares fitting.

[0100] Information and data related to spot coal prices include the closing price of the main thermal coal contract, the Baltic Dry Index (BDI), the China Coastal Coal Freight Index, coal freight rates from Qinhuangdao to Shanghai (4-5 DWT), coal freight rates from Qinhuangdao to Guangzhou (5-6 DWT), total coal inventory of the six coastal power plants, inventory of the four northern ports, inventory of Qinhuangdao Port, inventory of Guotou Jingtang Port, number of ships anchored at the four northern ports, inbound volume of the four northern ports, total coal consumption of the six coastal power plants, inventory of Guotou Caofeidian Port, and historical information on spot coal at the CCI spot index of 5500.

[0101] Artificial intelligence prediction methods are used to obtain spot coal market information data. This data is then filtered based on the characteristics of factors influencing spot coal prices. Historical data is input to train a spot coal price prediction model, which is then used to predict the trend of spot coal prices. An auxiliary database is constructed by coupling historical power generation information from generating units with future spot coal market information.

[0102] As an optional implementation, the least squares method required in step S1 includes:

[0103]

[0104] Among them, C t y represents the standard coal consumption for unit power generation during period t. t The historical standard coal consumption for power generation of the unit is given by f(t,a,b,c), which is a typical standard coal consumption curve function.

[0105] As an optional implementation, the spot coal price forecasting step required in step S1 includes:

[0106] Data processing: Cleaning and filling in the collected spot coal price-related data.

[0107] Model building: Using the processed data, a spot coal price prediction model is built. Historical data related to spot coal prices are input for model training to obtain network weights and training parameters.

[0108] Coal price forecast: Using a trained forecasting model combined with relevant information on spot coal prices, the system forecasts future spot coal prices and saves the forecast results.

[0109] As an optional implementation method, the spot coal price forecasting model includes:

[0110] A coal price prediction model based on the Stacking ensemble learning framework is used. By integrating intelligent prediction models such as PCA-LSTM, XGBoost, SVR, LSTM and RNN through the ensemble learning framework, swarm intelligence is achieved to improve prediction accuracy.

[0111] The ensemble learning framework includes a primary learner and a meta-learner, and employs K-flod cross-validation to achieve hierarchical coupling.

[0112] Selection of primary learners. The number of dissimilar points and the mean squared error in the predictions of each learner are used as measures to select primary learners with different differences, and they are combined into a model with high generalization performance.

[0113] Determine the meta-learner. The meta-learner must fully utilize the differences in predictions among the primary learners; the XGBoost model is used as the meta-learner. Introduce K-flod cross-validation. The cross-validation results of the first-layer primary learners are used as features for the training and prediction sets of the second-layer meta-learner.

[0114] S2. Based on the auxiliary database, design the minimum coal purchase / inventory cost function, calculate the minimum standard unit price for furnace input, and construct a two-level cost decision optimization model that measures the linkage between coal purchase and inventory in the spot market.

[0115] As an optional implementation method, a two-tiered cost decision optimization model linking coal purchase and inventory in the spot market is constructed, including:

[0116] Selecting coal suppliers by predicting coal prices, and choosing key indicators. The regulations are as follows:

[0117]

[0118] In the formula, and These are priority flags for purchasing spot coal and long-term contract coal, respectively. A value of 1 indicates that the m-period prioritizes the purchase of spot coal. A value of 0 indicates priority for the decomposition of long-term contract coal; λ m,j and These represent the predicted spot coal price for type j over period m and the average long-term contract coal price over a given period.

[0119] The upper-level model optimizes the procurement / inventory management decision based on the best choice between spot coal and long-term contract coal prices, with the total power generation cost as the objective, and transmits the coal purchase plan to the lower-level coal blending and combustion decision optimization; the total cost Y of the power generation enterprise's coal purchase decision is expressed as:

[0120]

[0121] In the formula, Let be the breakdown quantity (decision variable) of the j-th long-term coal contract in the m-th period. Let be the spot coal purchase volume in the m-th period (decision variable); Let j be the spot coal price in the m-th period. The price of the j-th long-term contract coal in the m-th cycle; S is the average warehouse management cost. m,i Let θ represent the inventory level for this period, and θ represent the unit inventory cost.

[0122] The lower-level model optimizes the coal blending scheme in the production process, calculates the unit cost of coal fed into the furnace, and transmits the coal consumption of each cycle to the upper-level coal procurement timing decision optimization; the coal blending ratio optimization model uses the lowest unit blending cost in each cycle. The objective function is:

[0123]

[0124] In the formula, it is assumed that ε m,i Let be the blending ratio of type i coal in the m-th cycle, then the unit blending cost in the m-th cycle is... The unit cost of fuel blending for power generation companies is determined by i; i represents the type of coal, B m,i Let p be the purchase quantity of the i-th type of coal in cycle m, and p be the coal delivery order. m,i , θ represents the average inventory management cost, and θ represents the unit inventory cost. The standard coal calorific value (7000 kcal), Q net,i The calorific value is that of ordinary coal;

[0125] The upper-level optimization aims to minimize coal purchase and inventory costs and then passes the coal purchase plan down to the lower level; the lower-level optimization uses the furnace feed list as the optimization objective to solve for the blending ratio and then passes the consumption of each type of coal up to the upper level.

[0126] S3. Based on the characteristics of the two-level cost decision optimization model, it is transformed into a single-level model using McCormick-KKT and solved using MATLAB to obtain the optimal cost of the base model.

[0127] As an optional implementation method, the two-level cost decision optimization model is characterized by the following: the upper-level coal purchasing decision model is a linear model, and its classification inventory constraints are influenced by the optimization results of the lower-level blending ratio optimization model. The objective function of the lower-level model contains bilinear terms, the form of which is indirectly determined by the solution results of the upper-level model. The two-level models influence each other, and the constructed two-level optimization model belongs to the two-level nonlinear optimization model.

[0128] As an optional implementation, step S3 specifically includes steps S31-S35:

[0129] S31. For bilinear terms in the lower-level model, replace the bilinear terms that cause nonlinearity with new variables.

[0130] S32. Add new constraints resulting from variable substitution.

[0131] S33. By combining the old and new constraints, the nonlinear terms containing bilinearity in the original lower-level optimization are transformed into linear terms.

[0132] S34. Use KKT conditions to solve for the extrema of the lower-level model, transforming the lower-level model into a series of constraints, thereby transforming the two-level nonlinear programming into a single-level linear programming.

[0133] S35. Solve a single-level integer linear programming model using MATLAB's CPLEX function.

[0134] As an optional implementation, the new constraints resulting from variable substitution in step S32 include:

[0135] According to the McCormick method, linearizing bilinear terms using variable substitution requires the introduction of the following new constraints:

[0136]

[0137] In the formula, x and y are the decision variables that are being replaced, and w is the decision variable that replaces the bilinear term of xy.

[0138] S4. Determine the expected cost based on the optimal cost of the base model, and construct an IGDT model to incorporate different risk attitudes in order to obtain risk decision optimization and robust decision optimization.

[0139] As an optional implementation, the constructed IGDT model includes:

[0140] The IGDT model based on robust optimization can be represented as:

[0141]

[0142] in, The robustness function (which becomes the opportunity-risk function when it reaches its minimum value) has an uncertainty factor α∈[0,1], determined by the coal purchase and blending decision scheme q(W,D,ε) and the expected cost threshold r. c Decision. R is the reward function, here set to the reciprocal of the cost of the thermal power plant, 1 / Y. The reward function is affected by the initial estimate of uncertainties. In this embodiment, the factor is the predicted spot coal price. U is defined as the set of all uncertain estimates u.

[0143] Using a fractional error model to create an extended interval for measuring future coal price volatility:

[0144]

[0145] S5. Based on risk decision optimization and robustness decision optimization, calculate the risk robustness-cost curve, and further revise the coal purchase / blending decision scheme according to actual preferences.

[0146] The above method will be explained in detail below with reference to the accompanying drawings and specific embodiments.

[0147] like Figure 1 As shown, this embodiment provides a two-level cost decision optimization method for thermal power plants based on IGDT (Information Technology for Determining Decision Risk), specifically including the following steps:

[0148] A1. Establish an auxiliary database. Obtain historical operating data of coal-fired power plants and use curve fitting methods to obtain the standard coal consumption characteristic curves for each unit; based on spot coal price information, obtain spot coal market information data through multiple spot coal price forecasting steps.

[0149] Table 1 contains historical operating data of thermal power units of power generation enterprises, including maximum output, minimum output, start-up cost, shutdown cost, shortest start-up time, shortest shutdown time, maximum ramp rate, and historical standard coal consumption for power supply. Each parameter name is replaced by λ1 to λ7 respectively.

[0150] Table 1 Historical Operating Data of Thermal Power Units

[0151]

[0152] like Figure 2 As shown, based on the above data, the standard coal consumption curve is fitted using the least squares method. The typical standard coal consumption curve function in this embodiment satisfies the following equation:

[0153] C t =0.0004t 2-0.4761t+456.79

[0154] In the formula, C t Let t represent the standard coal consumption for unit power generation during time period t. The coefficients are obtained by least squares fitting.

[0155] Least squares methods include:

[0156]

[0157] Among them, C t y represents the standard coal consumption for unit power generation during period t. t The historical standard coal consumption for power generation of the unit is given by f(t,a,b,c), which is a typical standard coal consumption curve function.

[0158] Information and data related to spot coal prices include the closing price of the main thermal coal contract, the Baltic Dry Index (BDI), the China Coastal Coal Freight Index, coal freight rates from Qinhuangdao to Shanghai (4-5 DWT), coal freight rates from Qinhuangdao to Guangzhou (5-6 DWT), total coal inventory of the six coastal power plants, inventory of the four northern ports, inventory of Qinhuangdao Port, inventory of Guotou Jingtang Port, number of ships anchored at the four northern ports, inbound volume of the four northern ports, total coal consumption of the six coastal power plants, inventory of Guotou Caofeidian Port, and historical information on spot coal at the CCI spot index of 5500.

[0159] Figure 3 This study utilizes artificial intelligence-based forecasting methods to obtain spot coal market data for April 2021, along with corresponding long-term contract coal prices. The results of forecasting spot coal price trends using this data are then used. An auxiliary database is constructed by coupling historical power generation information from generating units with future spot coal market information.

[0160] See Figure 4 The spot coal price prediction model includes:

[0161] A coal price prediction model based on the Stacking ensemble learning framework is used. By integrating intelligent prediction models such as PCA-LSTM, XGBoost, SVR, LSTM and RNN through the ensemble learning framework, swarm intelligence is achieved to improve prediction accuracy.

[0162] The ensemble learning framework includes a primary learner and a meta-learner, and employs K-flod cross-validation to achieve hierarchical coupling.

[0163] Selection of primary learners. The number of dissimilar points and the mean squared error in the predictions of each learner are used as measures to select primary learners with different differences, and they are combined into a model with high generalization performance.

[0164] Determine the meta-learner. The meta-learner must fully utilize the differences in predictions among the primary learners; the XGBoost model is used as the meta-learner. Introduce K-flod cross-validation. The cross-validation results of the first-layer primary learners are used as features for the training and prediction sets of the second-layer meta-learner.

[0165] A2. Based on the auxiliary database, design the minimum coal purchase / inventory cost function, calculate the minimum standard unit price for furnace input, and construct a two-level cost decision optimization model that measures the coal purchase / inventory linkage in the spot market.

[0166] A two-tiered cost decision optimization model linking coal purchase and inventory in the spot market is constructed. Supplier selection is based on spot coal price forecasts, forming an upper-level coal purchase / inventory decision optimization model and a lower-level blending ratio optimization model. Detailed steps include:

[0167] Selecting coal suppliers by predicting coal prices, and choosing key indicators. The regulations are as follows:

[0168]

[0169] In the formula, and These are priority flags for purchasing spot coal and long-term contract coal, respectively. A value of 1 indicates that the m-period prioritizes the purchase of spot coal. A value of 0 indicates priority for the decomposition of long-term contract coal; λ m,j and These represent the predicted spot coal price for type j over period m and the average long-term contract coal price over a given period.

[0170] (1) Upper-level model

[0171] The upper-level model optimizes procurement / inventory management decisions based on the best choice between spot coal and long-term contract coal prices, with the total power generation cost as the objective, and propagates the coal purchase plan to the lower-level blending coal combustion decision optimization. The total cost Y of the power generation company's coal purchase decision can be expressed as:

[0172]

[0173] In the formula, Let be the breakdown quantity (decision variable) of the j-th long-term coal contract in the m-th period. Let be the spot coal purchase volume in the m-th period (decision variable); Let j be the spot coal price in the m-th period. The price of the j-th type of long-term contract coal in the m-th cycle; S is the average warehouse management cost. m,i Let θ represent the inventory level for this period, and θ represent the unit inventory cost.

[0174] The constraints of the upper-level model are:

[0175] 1) Supplier supply constraints

[0176]

[0177] In the formula, for the j-th coal supplier, This is the minimum contract fulfillment volume for power generation companies. This is the upper limit for the amount that a supplier can provide.

[0178] 2) Categorized Coal Inventory Constraints

[0179]

[0180] S m,i =S m-1,i +B m,i -C m ε m,i

[0181] S min,i ≤S m,i ≤S max,i

[0182] In the formula, S m-1,i The inventory of type i coal in the previous procurement cycle, and type i coal in cycle m, B m,i Total purchase amount Let be the amount of coal decomposed in the j-th long-term contract during the m-th cycle. S represents the spot coal purchase volume in the m-th period. m,i C represents the corresponding total inventory. m Let ε be the total coal consumption in the m-th period. m,i The co-firing ratio for this period is input from the lower-level model; S min,t and S max,t These represent the upper and lower limits of coal storage for the i-th type, respectively.

[0183] 3) Total inventory constraints.

[0184]

[0185] In the formula, S min and S max These represent the upper and lower limits of the total power plant inventory, respectively.

[0186] (2) Lower-level model

[0187] The lower-level model optimizes the coal blending scheme in the production process, calculates the unit cost of coal fed into the furnace, and transmits the coal consumption of each cycle to the upper-level coal procurement timing decision optimization; the coal blending ratio optimization model uses the lowest unit blending cost in each cycle. The objective function is:

[0188]

[0189] In the formula, it is assumed that ε m,i Let be the blending ratio of type i coal in the m-th cycle, then the unit blending cost in the m-th cycle is... The unit cost of fuel blending for power generation companies is determined by i; i represents the type of coal, B m,i Let p be the purchase quantity of the i-th type of coal in cycle m, and p be the coal delivery order. m,i , θ represents the average inventory management cost, and θ represents the unit inventory cost. The standard coal calorific value (7000 kcal), Q net,i This is the calorific value of ordinary coal.

[0190] The constraints of the lower-level model are:

[0191] 1) Heating constraints

[0192]

[0193]

[0194] In the formula: G represents the standard coal consumption for the m-th cycle. m For the planned power generation in period m, g m Let η be the standard coal consumption per unit of generating electricity in the m-th cycle. g β represents the self-consumption rate of electricity generation, and β represents the boiler thermal efficiency.

[0195] 2) Inventory consumption constraints

[0196] 0≤C m ε m,i ≤S m-1,i +B m.i

[0197] 3) Production target constraints

[0198]

[0199]

[0200]

[0201] In the formula: L min and L max These represent the upper and lower limits of the average sulfur content in blended coal, respectively. i V represents the sulfur content of the i-th coal type; min and V max These are the upper and lower limits of volatile matter content on an ash-free basis for dried mixed coal, V i The dry ash-free volatile matter of coal type i; T min and Tmax These are the upper and lower limits of the ash melting point of mixed coal, T i Let be the ash melting point of the mixed coal of the i-th coal type.

[0202] The upper-level optimization aims to minimize coal purchase and inventory costs and then passes the coal purchase plan down to the lower level. The lower-level optimization uses the furnace feed order as the optimization objective to solve for the blending ratio and then passes the consumption of each type of coal up to the upper level.

[0203] A3. Based on the characteristics of the two-level cost decision optimization model, it is transformed into a single-level model through McCormick-KKT, and then solved using CPLEX in MATLAB to calculate the optimal cost of the base model.

[0204] The characteristics of the two-level cost decision optimization model include:

[0205] The upper-level coal purchasing decision model is a linear model, and its categorized inventory constraints are influenced by the optimization results of the lower-level blending ratio optimization model. The objective function of the lower-level model contains bilinear terms, the form of which is indirectly determined by the solution results of the upper-level model. The two models influence each other, and the constructed bi-level optimization model is a bi-level nonlinear optimization model.

[0206] See Figure 5 The steps for converting a two-layer model into a single-layer model and solving it include:

[0207] A31. For bilinear terms in the lower-level model, replace the bilinear terms that cause nonlinearity with new variables.

[0208] A32. Add new constraints resulting from variable substitution.

[0209] A33. By combining the old and new constraints, the nonlinear terms containing bilinearity in the original lower-level optimization are transformed into linear terms.

[0210] A34. Using KKT conditions to solve for the extrema of the lower-level model transforms the lower-level model into a series of constraints, thereby converting the two-level nonlinear programming into a single-level linear programming.

[0211] A35. Solve a single-level integer linear programming model using MATLAB's CPLEX function.

[0212] The new constraints resulting from variable substitution in step A32 include:

[0213] According to the McCormick method, after linearizing the bilinear terms using variable substitution and introducing new constraints, the objective function of the lower-level model is:

[0214]

[0215]

[0216] A4. Based on the optimal cost of the base model, set the expected cost and construct an IGDT model to incorporate different risk attitudes in order to obtain the decision risk to be tolerated and the robust optimization performance.

[0217] The general IGDT model based on risk aversion and robust optimization can be expressed as:

[0218]

[0219] in, The robustness function (which becomes the opportunity-risk function when it reaches its minimum value) has an uncertainty factor α∈[0,1], determined by the coal purchase and blending decision scheme q(W,D,ε) and the expected cost threshold r. c Decision. R is the reward function, here set to the reciprocal of the cost of the thermal power plant, 1 / Y. The reward function is affected by the initial estimate of uncertainties. In this paper, the factor is the predicted spot coal price. U is defined as the set of all uncertain estimates u. An extended interval quantity model is used to create the volatility of future coal prices using a fractional error model.

[0220]

[0221] In this embodiment, the risk-seeking opportunity model based on IGDT is as follows:

[0222] obj:minα op

[0223]

[0224] The reward function is set to the reciprocal of the cost of the thermal power plant, 1 / Y. Therefore, the lower-level objective function can be transformed into minimizing the cost Y to be less than the expected cost Y. c The expected cost is defined by the risk coefficient ε and the optimization result Y0 under the deterministic model.

[0225] In this embodiment, the risk avoidance model based on IGDT is as follows:

[0226] obj:maxα ro

[0227]

[0228] In this paper, the reward function is set as the reciprocal of the cost of the thermal power plant, 1 / Y. Therefore, constraint 1 can be transformed into the maximum cost being less than the expected cost Y. c The expected cost is defined by the robustness coefficient δ and the optimization result Y0 under the deterministic model.

[0229] A5. Calculate the risk robustness-cost curve based on risk decision optimization and robustness decision optimization, and further revise the coal purchase / blending decision scheme according to actual preferences, such as... Figure 9 As shown.

[0230] To verify the effectiveness of the method of this invention, a coal-fired power plant with an installed capacity of 2×660MW was taken as the research object. The parameter settings are shown in Table 1. Figures 2-4 This study explores indicators such as production cost, production inventory, and unit co-firing cost of thermal power plants.

[0231] Four scenarios are designed to analyze and compare the solutions of the classic single-layer model and the proposed two-layer model. Scenario 1: Solving the single-layer model based on the improved Grey Wolf Algorithm (IGWO); Scenario 2: Iteratively solving the two-layer model based on IGWO; Scenario 3: Solving the single-layer model based on Gurobi; Scenario 4: Solving the two-layer model using the KKT-McCormick method, but without incorporating the spot coal market.

[0232] Figure 6 To illustrate inventory changes under various scenarios, a comparison of single-level and two-level decision optimization models in terms of total inventory and category inventory was conducted for four scenarios. The results show that the total inventory of the IGWO-single-level model cannot be stably maintained at a safety stock level, and its inventory volatility is greater than that of the IGWO-two-level model. However, the two-level model based on the KKT-McCormick transformation, solved by a solver, yields the same total inventory result as the Gurobi-single-level model, effectively stabilizing inventory at a safety stock level. This not only satisfies the constraints but also minimizes inventory costs, demonstrating the effectiveness and stability of the proposed solution method.

[0233] Table 2 Comparison of Production Cost Indicators

[0234]

[0235] Figure 7 Table 2 compares the cost curves for scenarios 3 and 4. As shown in Table 2, the unit blending cost of the two-layer model is lower than that of the single-layer model for most of the time, with its average cost decreasing from 0.4986 yuan to 0.4761 yuan. The coal purchase cost and total cost are lower than those of the single-layer model throughout the entire period. The above analysis indicates that the two-layer model has a stronger coal purchase decision optimization capability. Secondly, the total cost of the two-layer decision optimization model is reduced from 77.7798 million yuan to 74.3976 million yuan compared to the single-layer model; the coal purchase cost is 63.1976 million yuan, a decrease of 3.3822 million yuan compared to the single-layer model. It is worth noting that scenario 4 is a two-layer decision optimization model based on KKT-McCormick, and its results are superior to the iterative solution results of the two-layer model based on IGWO, demonstrating the superiority and stability of the solution method and model proposed in this invention.

[0236] To explore the cost optimization effectiveness of the present invention, the two-level decision optimization model in scenario 4 was used as the control group, and scenario 5 was designed: a two-level decision optimization model, with the spot coal market introduced as the experimental group.

[0237] like Figure 8 As shown, Figure 8 For the comparison of coal purchase decisions under scenarios 4 and 5, on the 8th and 10th days, the predicted spot coal price is lower than the long-term contract coal price and is at a low level. By purchasing a large amount of spot coal to increase inventory, a large amount of spot coal is consumed on the 11th to 14th days when the spot coal price is relatively high, thereby achieving "increasing inventory at low levels, switching to power generation at high levels, and overall dynamic optimization".

[0238] Table 3 Comparison of Production Cost Indicators

[0239]

[0240] As shown in Table 3, the total cost decreased from RMB 74.3976 million to RMB 73.6529 million, which significantly reduced the power generation cost of the power plant. However, due to the slight increase in inventory costs and the significant decrease in coal purchase costs, the unit cost (the average unit co-firing cost) is smaller than that of the two-layer model.

[0241] As shown in Table 4, the two-level decision optimization model for thermal power plants based on IGDT is used to set different robustness coefficients δ, risk coefficients ε, and expected costs to obtain uncertainty measures and thermal power plant costs under different pursuit objectives.

[0242] Table 4. Uncertainty Measurement and Total Cost under Different Target Pursuits

[0243]

[0244] Table 4 presents the uncertainty factor measures and corresponding total power generation costs of thermal power plants under different robustness coefficients δ and risk coefficients ε. Under the risk-averse strategy, the estimated spot coal price will be greater than the predicted coal price, sacrificing some cost reduction potential in exchange for higher decision robustness. Taking a robustness coefficient δ = 0.4135% as an example, although the total cost of the optimization result of the two-level decision optimization model for thermal power plants based on IGDT is greater than that of the deterministic model, approximately 73.9575 million yuan, the uncertainty measure is 2.87%. That is, when the estimated deviation of the spot coal price is less than this value, the total cost will not be higher than the target value, thus providing decision-makers with higher decision stability.

[0245] In summary, compared with the prior art, the present invention has at least the following advantages and beneficial effects:

[0246] (1) This invention constructs a two-layer dynamic linkage decision optimization model for power generation inventory and production in the spot coal market by optimizing the upper layer of spot coal and long-term contract coal collaborative procurement / inventory management and optimizing the lower layer of coal blending and combustion scheme in the production process. This model achieves "low-level increase in inventory, high-level shift to power generation, and global dynamic optimization".

[0247] (2) The present invention converts the two-layer model into a single-layer model through the McCormick-KKT process, which has high solution efficiency and more accurate solution results.

[0248] (3) This invention measures the uncertainty risk of market coal price forecasts based on IGDT, sets different expected costs, and obtains robustness and risk measurement characteristics under different demands. It provides a more robust and less risky decision-making perspective for cost optimization of thermal power plants.

[0249] This embodiment also provides a two-layer cost decision optimization device for thermal power plants, including:

[0250] At least one processor;

[0251] At least one memory for storing at least one program;

[0252] When the at least one program is executed by the at least one processor, the at least one processor performs the following: Figure 1 The method shown.

[0253] This embodiment of a thermal power plant two-level cost decision optimization device can execute a thermal power plant two-level cost decision optimization method provided in the method embodiment of the present invention. It can execute any combination of implementation steps of the method embodiment and has the corresponding functions and beneficial effects of the method.

[0254] This application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform... Figure 1 The method shown.

[0255] This embodiment also provides a storage medium storing instructions or programs that can execute the two-level cost decision optimization method for thermal power plants provided in the method embodiment of the present invention. When the instructions or programs are run, any combination of implementation steps of the method embodiment can be executed, and the method has the corresponding functions and beneficial effects.

[0256] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0257] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0258] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0259] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0260] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0261] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0262] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0263] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0264] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A two-level cost decision optimization method for thermal power plants, characterized in that, Includes the following steps: S1. Establish an auxiliary database; wherein the auxiliary database includes coal consumption characteristic curves of thermal power units and spot coal market information data. S2. Based on the auxiliary database, design the minimum coal purchase / inventory cost function, calculate the minimum standard unit price for furnace input, and construct a two-level cost decision optimization model that measures the linkage between coal purchase and inventory in the spot market. S3. Based on the characteristics of the two-level cost decision optimization model, it is transformed into a single-level model through McCormick-KKT, and solved using MATLAB to obtain the optimal cost of the base model. S4. Determine the expected cost based on the optimal cost of the base model, and construct an IGDT model to incorporate different risk attitudes in order to obtain risk decision optimization and robust decision optimization. S5. Calculate the risk robustness-cost curve based on risk decision optimization and robustness decision optimization; The construction of a two-tier cost decision optimization model that links coal purchase and inventory in the spot market includes: Selecting coal suppliers by predicting coal prices, and choosing key indicators. The regulations are as follows: In the formula, and These are priority flags for purchasing spot coal and long-term contract coal, respectively. When it is 1, it means m Prioritize purchasing spot coal during the cycle. A value of 0 indicates that long-term contract coal will be prioritized for allocation; A value of 1 indicates that long-term contract coal will be prioritized for processing. A value of 0 indicates priority for purchasing spot coal; and They are respectively m cycle j Forecasted spot coal prices and average long-term contract coal prices over a period; The upper-level model optimizes procurement / inventory management decisions based on the best choice between spot coal and long-term contract coal prices, with the total power generation cost as the objective, and propagates the coal purchase plan to the lower-level coal blending and combustion decision optimization; the total cost of coal purchase decisions for power generation enterprises. Y The expression is as follows: In the formula, For the first j The long-term coal contract in the first m The amount of periodic decomposition, For the first m Periodic spot coal purchases; For the first m Cycle number j Spot coal prices, For the first m Cycle number j Long-term contract coal prices; To average warehouse management costs, This represents the inventory level for that period. θ Unit inventory cost; T The total number of cycles. M The number of coal suppliers Number of supplier categories; The lower-level model optimizes the coal blending scheme in the production process, calculates the unit cost of coal fed into the furnace, and transmits the coal consumption of each cycle to the upper-level coal procurement timing decision optimization; the coal blending ratio optimization model uses the lowest unit blending cost in each cycle. The objective function is: In the formula, it is assumed that For the first i Coal-fired power plants in the first m The blending ratio in the cycle, then the first m The cost per unit of co-firing in a cycle is This determines the unit cost of fuel blending for power generation companies; i Indicates the type of coal. for m Cycle number i The purchase volume of coal-like materials, and the coal delivery order for power plants are as follows: , To average warehouse management costs, θ Unit inventory cost The calorific value of standard coal. The calorific value is that of ordinary coal; The upper-level optimization aims to minimize coal purchase and inventory costs and then passes the coal purchase plan down to the lower level; the lower-level optimization uses the furnace feed list as the optimization objective to solve for the blending ratio and then passes the consumption of each type of coal up to the upper level.

2. The two-level cost decision optimization method for thermal power plants according to claim 1, characterized in that, The establishment of the auxiliary database includes: Historical operating data of thermal power units were obtained, and the coal consumption characteristic curves of each unit were obtained using curve fitting methods; Based on relevant information about spot coal prices, obtain spot coal market information data; train a spot coal price prediction model based on the spot coal market information data, and use the spot coal price prediction model to predict the trend of spot coal prices; Historical power generation information of thermal power units and future spot coal price market information are coupled to form an auxiliary database.

3. The two-level cost decision optimization method for thermal power plants according to claim 2, characterized in that, The historical operating data includes the maximum output, minimum output, start-up cost, shutdown cost, shortest start-up time, shortest shutdown time, maximum ramp rate, and historical standard coal consumption for power generation of the thermal power unit. The standard coal consumption curve of the thermal power unit is fitted using the least squares method, where the standard coal consumption curve function satisfies the following equation: In the formula, for t Standard coal consumption for power generation by thermal power units during a given period a , b , c These are the coefficients of the quadratic, linear, and constant terms of the standard coal consumption curve, respectively. The expression for the least squares method is as follows: In the formula, y t The historical standard coal consumption for power generation of thermal power units. f ( t , a , b , c ) is the standard coal consumption curve function.

4. The two-level cost decision optimization method for thermal power plants according to claim 2, characterized in that, The spot coal price prediction model is a coal price prediction model based on the Stacking ensemble learning framework. By integrating multiple intelligent prediction models through the ensemble learning framework, collective intelligence is achieved to improve prediction accuracy. The ensemble learning framework includes a primary learner and a meta-learner, and uses K-flod cross-validation to achieve hierarchical coupling. Selection of primary learners: The number of dissimilar points and the mean square error in the predictions of each learner are used as measures to select primary learners with different differences and combine them into a model with high generalization performance. Determine the meta-learner: Use the XGBoost model as the meta-learner and introduce K-flod cross-validation. Use the cross-validation calculation results of the first-layer primary learner as the features of the training set and prediction set of the second-layer meta-learner.

5. The two-level cost decision optimization method for thermal power plants according to claim 1, characterized in that, The features of the two-level cost decision optimization model include: The upper-level model is a linear model, and its classification inventory constraints are constrained by the optimization results of the lower-level blending ratio optimization model. The objective function of the lower-level model contains bilinear terms, the form of which is indirectly determined by the solution results of the upper-level model. The two-layer model influences each other, and the constructed two-layer optimization model belongs to the two-layer nonlinear optimization model.

6. The two-level cost decision optimization method for thermal power plants according to claim 5, characterized in that, Step S3 includes: S31. For bilinear terms in the lower-level model, replace the bilinear terms that cause nonlinearity with new variables; S32. Add new constraints resulting from variable substitution; S33. By combining the old and new constraints, the nonlinear terms containing bilinearity in the original lower-level optimization are transformed into linear terms. S34. Use KKT conditions to solve for the extrema of the lower-level model, transform the lower-level model into constraints, and thus transform the two-level nonlinear programming into a single-level linear programming. S35. Solve a single-level integer linear programming model using MATLAB's CPLEX function.

7. The two-level cost decision optimization method for thermal power plants according to claim 6, characterized in that, Step S32 includes: According to the McCormick method, linearizing bilinear terms using variable substitution requires the introduction of the following new constraints: In the formula, x and y The decision variable to be replaced. w For replacement xy Decision variables for bilinear terms; for m week i Coal-like decomposition amount for i Minimum blending ratio of coal types for i The maximum blending ratio of coal types. for m week i Minimum quantity of coal to be purchased for m week i The largest quantity of coal purchased.

8. The two-level cost decision optimization method for thermal power plants according to claim 1, characterized in that, The IGDT model constructed in step S4 includes: The robust optimization-based IGDT model is represented as follows: In the formula, Let be the robustness function, and let its value be the uncertainty factor. [0,1], determined by coal purchase and blending decision-making schemes q ( W, D, ε and expected cost threshold Decide; R Let the reward function be the initial estimate of the reward function, which is subject to uncertainty. Influence; U Defined as uncertainty estimation u The set of all ; Using a fractional error model to create an extended interval for measuring future coal price volatility: In the formula, This is the conversion factor.

9. A two-layer cost decision optimization device for thermal power plants, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method of any one of claims 1-8.

Images