An online real-time optimization method for cement calcination process based on time series prediction

Abstract

The present invention discloses an online real-time optimization method for cement calcining process based on time series prediction, characterized in that it comprises the following steps: Step 1: Collect key process parameters of cement calcining process for one month; Step 2: Based on the above key process parameters, construct a time series prediction model for cement calcining process; Step 3: Set the constraints of the process according to production requirements; Step 4: Take the minimization of heat loss in kiln tail exhaust gas as the optimization goal and construct the objective function; Step 5: Solve the optimization problem according to the optimization goal, constraints and time series prediction model to obtain the best adjustment strategy for the process. The present invention takes minimizing the production cost per ton of clinker as the goal, takes clinker quality index and calcining temperature as constraints, and performs online real-time optimization and adjustment on raw material feed amount, kiln head coal consumption and high temperature fan air volume in cement calcining process, so as to achieve continuous optimization of unit energy consumption under the premise of qualified quality and stable working conditions.

Description

Technical Field

[0001] The invention belongs to the field of cement production process modeling control and optimization, and relates to an online real-time optimization method for a cement calcination process based on time series prediction. Background Art

[0002] Fuel is one of the main costs in the cement production process, accounting for more than 40% of the total production cost. However, due to the extensive production operation mode of cement enterprises for a long time, the heat utilization efficiency of the calcination process is low, which will directly affect the operating costs of enterprises. Therefore, while ensuring stable production, maximizing fuel utilization and reducing enterprise production costs have become major problems that the industry needs to solve urgently.

[0003] Online real-time optimization (RTO) technology is based on historical production data and uses advanced AI modeling algorithms to establish a reliable twin model of the process, thereby guiding the system to automatically optimize combustion parameters in a complex and changeable industrial operating environment to ensure that the calcination process is efficient, stable, and low-emission. This type of method can not only significantly improve energy utilization efficiency and reduce operating costs, but also contribute to energy conservation and emission reduction and green and low-carbon transformation, help promote the modernization and upgrading of industrial production, and achieve a win-win situation for the economy and the environment.

[0004] The simplified process of cement calcination is as follows: Figure 2 The process flow diagram shown in the figure shows that the calcination unit includes a preheater, a decomposition furnace and a rotary kiln device, and the flue gas here corresponds to the flue gas at the outlet of the preheater. The input of the process includes raw materials, fuel, and air, among which the fuel may contain multiple types, such as coal powder, natural gas, garbage, and biomass alternative fuels. The air mainly provides the oxygen required for combustion; the output of the process includes the heated flue gas and the products formed by calcination, among which the main components of the flue gas are nitrogen, carbon dioxide, oxygen, carbon monoxide, etc.

[0005] Generally, in this production process, the means that can be adjusted through the distributed control system mainly include:

[0006] 1) The amount of air used: by adjusting the amount of air used, the fuel can be fully burned while taking into account the thermal efficiency of the system;

[0007] 2) The amount of fuel used: by adjusting the amount of fuel used, the stable calcination of raw materials and the quality of the products can be achieved;

[0008] 3) The amount of raw materials used. By adjusting the amount of raw materials used, both product quality and production cost can be taken into consideration.

[0009] The process adjustment objectives mainly include:

[0010] 1) Ensure stable operation of the equipment: It can be characterized by key process parameters. The common process indicator is the kiln head fire temperature;

[0011] 2) Ensure product quality: free calcium index of clinker products;

[0012] 3) Improve the economic benefits of the process: Set different production targets based on the market demand for the product, such as maximizing production capacity when market demand is strong; and reducing unit production costs when market demand is low, where unit production costs mainly include fuel costs and electricity costs.

[0013] At present, the application of intelligent control optimization of calcination process is mainly concentrated in the field of advanced process control, and its main goal is to stabilize the key process parameters of the process, such as using PID, MPC and other control algorithms to control the calcination temperature by adjusting the amount of raw materials and fuel; by adjusting the amount of air, the residual oxygen concentration at the outlet of the calcination process is controlled; this solution can effectively ensure the stable operation of the process, and even in some processes where the corresponding relationship between calcination temperature and product quality is clear, it can ensure the stability of calcination product quality. However, the solution of dividing and conquering wind, fuel, and raw materials lacks consideration of the overall economic benefits of the system. For example, when the fuel and wind consumption are high, the heat in the fuel may not be fully utilized by the raw materials due to insufficient combustion time and low gas-solid heat exchange efficiency, resulting in a waste of energy costs.

[0014] In addition, traditional methods will guide advanced control systems to make more energy-saving decisions based on the above-mentioned advanced control, with the sum of raw material, fuel, and electricity consumption costs as the target. However, current research mainly guides optimization by minimizing the following objective function:

[0015]

[0016] in, Q p Indicates the yield of the product, in t / h; Q f Indicates the flow rate of fuel, in t / h, C f It represents the unit cost of fuel, in Yuan / t; P f Indicates the power of the fan, in kW. C e Represents the electricity cost, in units of RMB / kWh. However, due to the calorific value of the fuel, inaccurate measurement and other reasons, the above models often cause mismatches between the objective function and the real problem, resulting in abnormal solution results; in addition, the decision results directly based on the objective function tend to increase the system's production, reduce fuel, and reduce wind. Therefore, this solution has extremely high requirements for the modeling of the calcination process constraints, and in actual applications, it often cannot be implemented because the accuracy cannot meet the requirements. Summary of the invention

[0017] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to propose an online real-time optimization method for cement calcination process based on time series prediction.

[0018] The technical solution of the present invention to achieve its technical purpose is as follows:

[0019] An online real-time optimization method for cement calcination process based on time series prediction comprises the following steps:

[0020] Step 1: Collect key process parameters of cement calcination process for one month;

[0021] Step 2: Based on the above key process parameters, a time series prediction model of cement calcination process is constructed;

[0022] Step 3: Set process constraints according to production requirements;

[0023] Step 4: Taking the minimization of heat loss in the kiln exhaust gas as the optimization goal, construct the objective function;

[0024] Step 5: Solve the optimization problem according to the optimization goal, constraints and time series prediction model to obtain the best adjustment strategy for the process.

[0025] The key process parameters described in step 2 include the coal feeding amount at the kiln tail, the amount of alternative fuel, the decomposition furnace temperature, the coal feeding amount at the kiln head, the raw material feeding amount, the air volume of the high-temperature fan, the O2 concentration in the rotary kiln smoke chamber, the CO concentration in the rotary kiln smoke chamber, the CO concentration at the decomposition furnace outlet, the O2 concentration at the decomposition furnace outlet, the kiln head fire-watching temperature, the preheater outlet temperature, the preheater outlet CO concentration, and the preheater outlet O2 concentration.

[0026] The time series prediction model described in step 2 has two input parts: 1) historical process production data, including coal feeding amount at the kiln tail, alternative fuel usage, decomposition furnace temperature, coal feeding amount at the kiln head, raw meal feeding amount, high-temperature fan air volume, O2 concentration in the smoke chamber of the rotary kiln, CO concentration in the smoke chamber of the rotary kiln, CO concentration at the outlet of the decomposition furnace, O2 concentration at the outlet of the decomposition furnace, fire-watching temperature at the kiln head, preheater outlet temperature, CO concentration at the outlet of the preheater, and O2 concentration at the outlet of the preheater; 2) future adjustment strategies for raw meal flow, coal feeding amount at the kiln head, and high-temperature fan air volume; the output of the model is the key process parameters related to control optimization in the future time, including fire-watching temperature at the kiln head, CO concentration in the smoke chamber, O2 concentration at the outlet of the decomposition furnace, preheater outlet temperature, and CO concentration at the outlet of the preheater.

[0027] In the time series prediction model, the time series prediction result is obtained by adding the open-loop prediction sequence and the future action response sequence.

[0028] The open-loop prediction sequence part includes an Embedding layer, an Encoding layer and a Decoding layer, wherein the Encoding layer outputs a vectorized state description of the system at the current moment, and the Decoding layer decodes and outputs the change trend of the process parameters in the future without active adjustment.

[0029] The input of the future action response part includes the state description and future action sequence obtained by the Encoding layer.

[0030] The constraints of the setting process described in step 3 include a lower limit constraint on the clinker calcining temperature, an upper limit constraint on the CO concentration in the smoke chamber, and a lower limit constraint on the O2 concentration at the outlet of the decomposition furnace.

[0031] The objective function described in step 4 is the unit production cost of clinker, and the cost is composed of the fan electricity cost and the fuel cost; the fan electricity cost is obtained by multiplying the fan power per unit output and the real-time electricity price; the fuel cost is obtained by multiplying the unit calorific value cost and the heat consumed per unit output of clinker, wherein the unit calorific value cost is obtained by weighting the current fuel consumption per ton of clinker, the fuel calorific value, and the unit price; the heat consumed per unit output of clinker is the heat loss carried away by the exhaust gas at the outlet of the preheater, including the sensible heat of the exhaust gas and the chemical heat contained in the CO that is not fully burned in the exhaust gas.

[0032] The optimization problem described in step 5 is solved, and the optimization target consists of multiple parts, including the clinker unit production cost and the system control stability constraint penalty item; the decision variables in the optimization decision stage are the adjustment sequence of the head coal consumption, the high-temperature fan speed, and the feed amount in the future time.

[0033] Beneficial effects of the present invention:

[0034] The present invention aims to construct a unified calcination prediction model for the preheater, decomposition furnace and rotary kiln in the cement calcination process, and to minimize the production cost per ton of clinker as the goal, with clinker quality indicators and calcination temperature as constraints, to perform online real-time optimization and adjustment of the raw material feed amount, kiln head coal consumption, and high-temperature fan air volume in the cement calcination process, so as to achieve continuous optimization of unit energy consumption under the premise of qualified quality and stable operating conditions.

[0035] The present invention can be applied to the domestic dry cement calcination production line. The system reduces the heat loss cost of the exhaust gas discharged from the preheater outlet by adjusting the kiln head coal feeding, high-temperature fan speed, and raw material feeding amount, and can reduce the fuel cost of the production line by more than 1.5%. It is expected to save more than 1.5 million yuan in production costs per year for a standard 5000t production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a structural diagram of a cement calcination process timing prediction model presented in the present invention.

[0037] Figure 2 It is a simplified process flow diagram of cement calcination process. DETAILED DESCRIPTION

[0038] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings. It should be noted that the various symbols in the expressions involved in the present invention are only for simplification and clear expression, and different expressions can be used in different examples. Therefore, the key to the present invention lies in the construction method of the objective function for optimizing the cement calcination process and the structure of the process time series prediction model.

[0039] The present invention aims to propose an online real-time optimization method for cement calcination process based on time series prediction, which includes the following five steps:

[0040] Step 1: Collect key process parameters of cement calcination process for about one month;

[0041] Step 2: Based on the above production data, a time series prediction model for the cement calcination process is constructed;

[0042] Step 3: Set process constraints according to production requirements;

[0043] Step 4: Construct an objective function with the goal of minimizing the heat loss in the kiln exhaust gas;

[0044] Step 5: According to the above optimization objectives, constraints and process models, solve the optimization problem and obtain the best adjustment strategy for the process.

[0045] In step 1, the most critical process parameters in the cement calcination process need to be collected, including the amount of coal fed to the kiln tail, the amount of alternative fuels, the temperature of the decomposition furnace, the amount of coal fed to the kiln head, the amount of raw material fed, the air volume of the high-temperature fan, the O2 concentration in the smoke chamber of the rotary kiln, the CO concentration in the smoke chamber of the rotary kiln, the CO concentration at the outlet of the decomposition furnace, the O2 concentration at the outlet of the decomposition furnace, the fire-watching temperature at the kiln head, the outlet temperature of the preheater, the CO concentration at the outlet of the preheater, and the O2 concentration at the outlet of the preheater. The time length is not less than one month, and the sampling period is not more than 10s. It is further necessary to pre-process the original data according to the characteristics of the production device, including eliminating the shutdown data, eliminating the instrument failure data, and filtering the gas composition instrument with the back-blowing process to ensure that all process data can truly reflect the working characteristics of the process.

[0046] In step 2, first construct a training sample data set according to the requirements of the collected data< s,u,y >, where All process parameters in the past period of time, that is, all process parameters collected in step 1; It indicates the operation sequence of kiln head coal feeding amount, high temperature fan speed and raw material feeding amount in the future period; It indicates the change sequence of key process parameters in the future, including the kiln head fire temperature, smoke chamber CO concentration, decomposition furnace outlet O2 concentration, and preheater outlet temperature. s,u,y >, for a model structure such as Figure 1 The model shown in Figure 2 has the following input: s,u The output of the model is its predicted output result , then the loss function in the training process can use the MSE indicator between the predicted results and the true labels.

[0047] The structure of the time series prediction model is as follows Figure 1 As shown in the figure, the input of the model includes two parts: 1) historical process production data; 2) future adjustment strategies for raw material flow, kiln head coal feeding amount, and high-temperature fan air volume. The model is divided into two parts, the left part is the open-loop prediction result based on historical data, that is, the change trend of the system output variables without any adjustment in the future. This part mainly includes three layers. The first layer is the Embedding layer of the process variables, which aims to compress and extract low-dimensional features of the variables; the second layer is the Encoding layer of the process state, which can use the Encoding part of the commonly used seq2seq model, which extracts high-dimensional features from historical data to form a vectorized description of the current process state; the third layer of the Decoding layer decodes the state information to obtain the open-loop time series prediction result of the process. The right part mainly realizes the prediction results under different future decision-making operations. It mainly includes a two-layer model structure. The first layer realizes the embedding of different operation records, such as evaluating the matching relationship between fuel and wind; the second layer realizes the superposition of the impact on future operation sequences. The dynamic characteristics of its impact are affected by the process state representation vector on the left, thereby characterizing the nonlinear relationship between the operating variable to the output variable under different working conditions.

[0048] At this point, the time series prediction model constructs the changing trend of future key process parameters under different working conditions and different future operation decisions of the calcination process, that is, the model shown below is obtained:

[0049]

[0050] The observed variables of the process are S krepresents all relevant variables that can be detected by measuring instruments at time k, including the amount of coal fed to the kiln tail, the amount of alternative fuel, the temperature of the decomposition furnace, the amount of coal fed to the kiln head, the amount of raw meal fed, the air volume of the high-temperature fan, the O2 concentration in the smoke chamber of the rotary kiln, the CO concentration in the smoke chamber of the rotary kiln, the CO concentration at the outlet of the decomposition furnace, the O2 concentration at the outlet of the decomposition furnace, the fire-watching temperature at the kiln head, the outlet temperature of the preheater, the CO concentration at the outlet of the preheater, and the O2 concentration at the outlet of the preheater; the operating variables of the process U k Indicates the variables that can be directly manipulated at time k, including the coal consumption, raw material consumption, and high-temperature fan air volume of each kiln head; the key output variables of the process y k represents the key process parameters related to control optimization at time k, including kiln head fire temperature, smoke chamber CO concentration, decomposition furnace outlet O2 concentration, preheater outlet temperature, and preheater outlet CO concentration; P represents the length of the prediction sequence, T represents the length of historical data required to describe the process state, and M represents the length of future operating variables taken into account.

[0051] Steps 3 and 4 mainly construct the objective function and constraints of the optimization problem. The constraints mainly include the lower limit constraint on the clinker calcination temperature, the upper limit constraint on the CO concentration in the smoke chamber, and the lower limit constraint on the O2 concentration at the outlet of the decomposition furnace. The clinker calcination temperature can be selected as a reference for the kiln head fire temperature, the CO concentration in the smoke chamber indicates the ventilation status in the kiln, and the lower limit constraint on the O2 concentration at the outlet of the decomposition furnace ensures that the tail coal is fully burned in the decomposition furnace. Based on this, the first part of the control optimization decision, Loss, is obtained:

[0052]

[0053] in, is the upper control limit of the i-th dimension constraint variable, is the lower control limit of the i-th dimension constraint variable, represents the prediction result of the i-th dimension constraint variable at time k+j, λ i represents the weight factor of the i-th dimension constraint variable, is the relaxation factor of the i-th dimension constraint variable at time k+j, n y =3 indicates the dimension of the constraint variable.

[0054] The objective function in step 4 represents the heat loss taken away by the exhaust gas at the outlet of the preheater, which mainly includes the sensible heat of the exhaust gas and the chemical heat contained in the CO that is not fully burned in the exhaust gas. The calculation method of the heat loss per unit of the process is as follows:

[0055]

[0056] in Q k+jIndicates the flue gas volume at time k+j, in Nm 3 / h, T g , k+j Indicates flue gas temperature, CO k+j represents the concentration of CO, γ CO Indicates the heat of combustion of CO in kJ / Nm 3 , ρ g Indicates the density of the outlet smoke, c g represents the specific heat capacity, Q p , k+j It represents the yield of the product at time k+j, in t / h. H c Represents the heat loss per unit, in kJ / t. So far, we have calculated the heat loss per unit, which mainly corresponds to the fuel cost of the process. In this solution, we first calculate the unit calorific value cost:

[0057]

[0058] in, C f Indicates the cost per unit calorific value, in yuan / kJ, Indicates the current consumption of the i-th fuel, in t / h, Indicates its unit cost, unit yuan / t, represents the calorific value of the i-th fuel, in kJ / t. Based on the above results, we further obtain the unit production cost:

[0059]

[0060] in, J Indicates the clinker cost per unit output, unit: RMB / t. Indicates the high temperature fan power at time k+j, in kW, is the real-time electricity price, in RMB / kWh, Q p , k+j It represents the product yield at time k+j, in t / h.

[0061] So far, the problem to be optimized in step 5 can be obtained:

[0062]

[0063] The optimization objective function consists of two parts. The first half represents the economic goal, and the second half represents the penalty term constrained by the production stability index. The decision variables include the operation sequence of the three main adjustment variables in the future M steps.

[0064] For the above optimization problems, conventional optimization algorithms such as grid search method, random search method, Bayesian optimization algorithm, particle swarm algorithm, etc. can be used to solve them.

[0065] In one implementation case of the present invention, the number of decision steps M for the above problem is 1, that is, only one action in the future is planned, thereby greatly reducing the complexity of solving the problem; a grid search method is used to search for the best decision action, that is, within a given upper and lower limit range of the operation, the solution space is divided into equally spaced grids according to a given solution accuracy, and each point in the grid corresponds to an adjustment strategy. All feasible solutions are recorded as , input the decision plan and historical status data into the model established in step 2 to obtain the model prediction result , and finally the prediction results of the model Substitute into the objective function constructed in steps 3 and 4 to obtain the evaluation index , we sort the above feasible solutions according to the evaluation indicators and get the optimal solution: Finally, the system writes the optimal solution calculated every minute, i.e. the optimal kiln head coal feeding amount, raw material feeding amount and high-temperature fan speed, back to the DCS system to realize automatic closed-loop adjustment of on-site equipment.

[0066] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, all of which belong to the protection scope of the present invention. The protection scope of the present invention is given by the attached claims and any equivalent technical solutions thereof.

Claims

1. An online real-time optimization method for cement calcination process based on time series prediction, characterized in that: The steps include: Step 1: Collect key process parameters of cement calcination process; Step 2: Based on the above key process parameters, a time series prediction model of cement calcination process is constructed; Step 3: Set process constraints according to production requirements; Step 4: Taking the minimization of heat loss in the kiln exhaust gas as the optimization goal, construct the objective function; Step 5: Solve the optimization problem according to the optimization goal, constraint conditions and time series prediction model to obtain the best adjustment strategy for the process; The key process parameters described in step 1 include the amount of coal fed to the kiln tail, the amount of alternative fuel, the temperature of the decomposition furnace, the amount of coal fed to the kiln head, the amount of raw material fed, the air volume of the high-temperature fan, the O2 concentration in the smoke chamber of the rotary kiln, the CO concentration in the smoke chamber of the rotary kiln, the CO concentration at the outlet of the decomposition furnace, the O2 concentration at the outlet of the decomposition furnace, the fire-watching temperature at the kiln head, the outlet temperature of the preheater, the CO concentration at the outlet of the preheater, and the O2 concentration at the outlet of the preheater; For the key process parameters, the collection time is not less than one month, and the sampling period is not longer than 10s. For the data of the key process parameters, the original data is pre-processed according to the characteristics of the production equipment, including eliminating the shutdown data, eliminating the instrument failure data, and filtering the gas composition instrument with backwash process to ensure that all process data can truly reflect the working characteristics of the process; The time series prediction model described in step 2 has two input parts: 1) historical process production data, including kiln tail coal feeding amount, alternative fuel usage, decomposition furnace temperature, kiln head coal feeding amount, raw meal feeding amount, high-temperature fan air volume, rotary kiln smoke chamber O2 concentration, rotary kiln smoke chamber CO concentration, decomposition furnace outlet CO concentration, decomposition furnace outlet O2 concentration, kiln head fire-watching temperature, preheater outlet temperature, preheater outlet CO concentration, preheater outlet O2 concentration; 2) future adjustment strategy for raw meal flow, kiln head coal feeding amount, and high-temperature fan air volume; the output of the model is the key process parameters related to control optimization in the future time, including kiln head fire-watching temperature, smoke chamber CO concentration, decomposition furnace outlet O2 concentration, preheater outlet temperature, and preheater outlet CO concentration; The time series prediction model is divided into two parts, the left part is the open-loop prediction result based on historical data, which includes three layers. The first layer is the Embedding layer of the process variables, which compresses and extracts low-dimensional features of the variables; The second layer is the encoding layer of the process state, which uses the encoding part of the seq2seq model to extract high-dimensional features from historical data to form a vectorized description of the current process state; the third layer of the decoding layer decodes the information of the state to obtain the open-loop time series prediction results of the process; the right part realizes the prediction results under different future decision operations, including a two-layer model structure. The first layer realizes the embedding of different operation records, and the second layer realizes the superposition of the impact on the future operation sequence. The dynamic characteristics of the impact are affected by the process state representation vector on the left, so as to characterize the nonlinear relationship between the operating variable and the output variable under different working conditions; The expression of the time series prediction model is: ; The observed variables of the process are S k represents all relevant variables that can be detected by measuring instruments at time k, including the amount of coal fed to the kiln tail, the amount of alternative fuel, the temperature of the decomposition furnace, the amount of coal fed to the kiln head, the amount of raw meal fed, the air volume of the high-temperature fan, the O2 concentration in the smoke chamber of the rotary kiln, the CO concentration in the smoke chamber of the rotary kiln, the CO concentration at the outlet of the decomposition furnace, the O2 concentration at the outlet of the decomposition furnace, the fire-watching temperature at the kiln head, the outlet temperature of the preheater, the CO concentration at the outlet of the preheater, and the O2 concentration at the outlet of the preheater; the operating variables of the process U k Indicates the variables that can be directly manipulated at time k, including the coal consumption, raw material consumption, and high-temperature fan air volume of each kiln head; the key output variables of the process y k represents the key process parameters related to control optimization at time k, including kiln head fire temperature, smoke chamber CO concentration, decomposition furnace outlet O2 concentration, preheater outlet temperature, and preheater outlet CO concentration; P represents the length of the prediction sequence, T represents the length of historical data required for process state description, and M represents the length of future operating variables considered; The constraints of the setting process described in step 3 include the lower limit constraint on the clinker calcining temperature, the upper limit constraint on the CO concentration in the smoke chamber, and the lower limit constraint on the O2 concentration at the outlet of the decomposition furnace, and the expression is: ； in, L 1 The first part of the control optimization decision is Loss, is the upper control limit of the i-th dimension constraint variable, is the lower control limit of the i-th dimension constraint variable, represents the prediction result of the i-th dimension constraint variable at time k+j, λ i represents the weight factor of the i-th dimension constraint variable, is the relaxation factor of the i-th dimension constraint variable at time k+j, n y =3 indicates the dimension of the constraint variable; The objective function described in step 4 is the unit production cost of clinker, which is composed of the electricity cost of the fan and the fuel cost; the electricity cost of the fan is obtained by multiplying the unit fan power and the real-time electricity price; the fuel cost is obtained by multiplying the unit calorific value cost and the heat consumed by the unit clinker production, wherein the unit calorific value cost is obtained by weighting the current fuel consumption per ton of clinker, the fuel calorific value, and the unit price; the heat consumed by the unit clinker production is the heat loss taken away by the exhaust gas at the outlet of the preheater, including the sensible heat of the exhaust gas and the chemical heat contained in the CO that is not fully burned in the exhaust gas; The unit calorific value cost is as follows: ； in, C f Indicates the cost per unit calorific value, in yuan / kJ, Indicates the current consumption of the i-th fuel, in t / h, Indicates its unit cost, unit yuan / t, Indicates the calorific value of the i-th fuel, in kJ / t; The calculation method of heat consumption per unit of clinker production is as follows: ； in Q k+j Indicates the flue gas volume at time k+j, in Nm 3 / h, T g , k+j Indicates flue gas temperature, CO k+j represents the concentration of CO, γ CO Indicates the heat of combustion of CO in kJ / Nm 3 , ρ g Indicates the density of the outlet smoke, c g represents the specific heat capacity, Q p , k+j It represents the yield of the product at time k+j, in t / h. H c Indicates the heat loss per unit, in kJ / t; The unit production cost of clinker is as follows: ； in, J Indicates the unit clinker production cost, unit yuan / t, P f , k+j Indicates the high temperature fan power at time k+j, in kW, C e is the real-time electricity price, in RMB / kWh, Q p , k+j It represents the product yield at time k+j, in t / h; Based on the unit calorific value cost, the heat consumed per unit of clinker production and the unit clinker production cost, the expression of the problem to be optimized in step 5 is as follows: 。 2. The method according to claim 1, characterized in that For the problem to be optimized and solved in step 5, a grid search method, a random search method, a Bayesian optimization algorithm, or a particle swarm algorithm is used to solve it; during the solving process, the number of decision steps of the problem is set to 1 to reduce the complexity of solving the problem.

Images