A coal-fired power plant operation parameter self-adaptive optimization method and system

Abstract

The application discloses a kind of coal-fired power plant operating parameter adaptive optimization method and system.The method includes operating data fusion acquisition, dynamic load response modeling, coal load coordination optimization, emission efficiency adaptive balance and operating parameter adaptive optimization.The application relates to the technical field of coal-fired power plant parameter intelligent adjustment, by constructing standardization multi-source data structure, using long short-term memory network to realize future load minute level rolling prediction, and introducing adaptive two-stage collaborative scheduling and improved multi-objective optimization algorithm to realize combustion parameter dynamic generation.On this basis, the rolling weighted balance of emission efficiency is carried out by using dynamic causal control chart, the energy efficiency, responsiveness and emission compliance are comprehensively considered, the power generation efficiency is improved, the pollution emission is reduced, and the system adaptability and stability are enhanced.

Description

Technical Field

[0001] This invention relates to the field of intelligent parameter adjustment technology for coal-fired power plants, specifically to an adaptive optimization method and system for operating parameters of coal-fired power plants. Background Technology

[0002] The adaptive optimization method and system for operating parameters of coal-fired power plants is a technology based on real-time data-driven and intelligent algorithms. It dynamically adjusts control strategies by monitoring the operating parameters (such as temperature, pressure, and coal consumption) of key equipment like boilers and turbines online, combined with machine learning or optimization algorithms, ensuring the power plant is always operating at its optimal condition. Its function is to automatically adapt to dynamic conditions such as coal quality fluctuations and load changes, reducing coal consumption and pollutant emissions, improving power generation efficiency, and minimizing human intervention, thus achieving intelligent closed-loop operation control.

[0003] However, in the existing adaptive optimization of operating parameters of coal-fired power plants, there are technical problems such as coarse granularity of multi-source data fusion, low accuracy of load prediction and rigid regulation response mechanism. This makes it difficult for the control system to respond effectively in a timely manner when there are drastic load fluctuations or drastic changes in coal quality.

[0004] Existing methods for coordinating and optimizing coal-fired loads have technical problems such as insufficient identification of key adjustment variables and deviation of parameter configuration results from system stability constraints. In particular, in multi-variable linkage adjustment scenarios, traditional methods cannot take into account the complex coupling relationship between load response, energy efficiency and emission targets.

[0005] Existing emission efficiency adaptation balance methods suffer from the technical problem of lacking the ability to dynamically model the causal relationship between adjustment parameters and pollutant emissions, leading to lags in strategy adjustments and frequent exceedances of pollutant standards. Summary of the Invention

[0006] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an adaptive optimization method and system for operating parameters of coal-fired power plants. Addressing the technical problems of coarse-grained multi-source data fusion, low load prediction accuracy, and rigid regulation response mechanisms in existing adaptive optimization methods for coal-fired power plant operating parameters, which lead to difficulties in timely and effective responses from the control system under conditions of drastic load fluctuations or coal quality changes, this solution creatively integrates three functional modules: dynamic load response modeling, coal-fired load coordination optimization, and emission efficiency adaptive balancing. This constructs a multi-module coupled closed-loop adaptive regulation link, enabling accurate fusion processing of multi-dimensional operating data, minute-level prediction of future loads, and dynamic optimization configuration of control parameters in actual operating conditions. Furthermore, addressing the technical problems of insufficient identification of key regulating variables and deviations of parameter configuration results from system stability constraints in existing coal-fired load coordination optimization methods, especially in multi-variable linkage regulation fields, this invention also addresses these issues. In this context, traditional methods cannot simultaneously address the complex coupling relationship between load response, energy efficiency, and emission targets. This solution creatively introduces an adaptive two-stage collaborative scheduling method. First, a linear mapping is used to achieve coarse-grained configuration, followed by a multi-objective optimization algorithm for nonlinear refinement. Finally, simulation verification confirms the construction of a stable parameter set, effectively improving the feasibility and engineering practicality of parameter generation. Addressing the technical problem in existing emission efficiency adaptive balancing methods—namely, the lack of dynamic modeling capability for the causal relationship between adjustment parameters and pollutant emissions, leading to lags in strategy adjustments and frequent exceedances of pollutant standards—this solution creatively introduces a multi-constraint rolling weighted balancing method based on a dynamic causal control graph. This constructs a causal feedback model guided by emission targets and incorporating sensitivity scoring, and designs a dynamic priority adjustment mechanism with a rolling window. This achieves a combination of emission level assessment, adjustment command optimization, and strategy selection, enhancing the foresight and compliance assurance capabilities of strategy adjustment.

[0007] The technical solution adopted by this invention is as follows: This invention provides an adaptive optimization method for operating parameters of a coal-fired power plant, which includes the following steps:

[0008] Step S1: Run data fusion and acquisition;

[0009] Step S2: Dynamic load response modeling;

[0010] Step S3: Coal-fired load coordination optimization;

[0011] Step S4: Emission efficiency adaptation balance;

[0012] Step S5: Adaptive optimization of runtime parameters.

[0013] Furthermore, in step S1, the operation data fusion and acquisition is used to collect and fuse multi-dimensional data on the operation status of coal-fired power plants under different operating conditions. Specifically, it involves collecting coal-fired power plant operation parameter data through multi-source data acquisition, including coal characteristic data, real-time operating condition data, control system parameters, historical operating load data, and environmental emission detection data.

[0014] By performing unified timestamp alignment, missing value imputation, outlier removal, feature standardization, and category mapping on the operating parameter data of coal-fired power plants, and constructing a standardized operating status vector sequence based on a preset field structure, the fused multi-source operating data of coal-fired power plants is obtained for subsequent predictive modeling and optimization analysis.

[0015] The coal-fired power plant operates using multi-source data, which is a sequence of feature vectors with a unified structure.

[0016] Furthermore, in step S2, the dynamic load response modeling is used to construct a load response prediction model for coal-fired power plants under different operating conditions and control strategies, thereby improving the foresight and scheduling adaptability of load regulation. Specifically, based on the multi-source operating data of the coal-fired power plant, a standard long short-term memory network is used to perform time modeling on the historical load sequence, and the current operating condition characteristics are combined as input to obtain the predicted load sequence data.

[0017] Further, in step S3, the coal-fired load coordination optimization, based on the predicted load trend, involves the coordinated optimization configuration of various operating sub-parameters in the coal-fired power generation system to achieve a dynamic balance between combustion efficiency, load response speed, and emission control. Specifically, based on the predicted load sequence data and multi-source operating data of the coal-fired power plant, an adaptive two-stage coordinated scheduling method is used to perform coal-fired load coordination optimization to obtain initial combustion configuration parameters, including the following steps:

[0018] Step S31: Identification of key load regulation variables, used to identify the regulation variables most sensitive to predicted load changes from multi-source input features. Specifically, based on coal characteristic data, control system parameters and predicted load sequence data, a feature importance assessment model is constructed. An improved Shapley value method and cross-information gain fusion strategy are used to extract a set of key regulation parameters that have a high response to load changes, including air supply volume, primary and secondary air ratio, coal blending ratio and burner channel status.

[0019] Step S32: Initial adjustment parameter generation, used to construct the first stage coarse-grained adjustment parameter configuration while maintaining the tracking of the predicted load. Specifically, it constructs a linear regression mapping relationship based on constrained least squares to map the predicted load sequence to the adjustment variable space, and introduces the historical operating load data to form upper and lower limit constraints to obtain the initial vector of combustion initial settings, including the initial air supply volume, primary and secondary air ratio and coal type ratio vector.

[0020] Step S33: Multi-objective collaborative refinement optimization, used to perform nonlinear multi-objective refinement adjustment for the three major objectives of energy saving, environmental protection and response rate. Specifically, it constructs a multi-objective optimization algorithm based on improved differential evolution, with the objective functions of maximizing combustion efficiency, minimizing unit load emissions and minimizing response delay. It integrates the predicted load gradient and operating dynamic conditions to iteratively optimize the initial parameters and generate the optimal configuration parameters that meet the objective constraints.

[0021] Step S34: Parameter stability optimization, used to dynamically simulate and verify the stability of the output parameters and the feasibility of the system under actual working conditions. Specifically, the optimized configuration parameters are input into the simulation module based on physical model constraints to perform thermal balance, air-coal flow field simulation and emission prediction. If the output indicators meet the safety boundary and response requirements, the configuration is confirmed as the initial combustion configuration parameters. Otherwise, step S33 is repeated for reconstruction iteration to obtain the initial combustion configuration parameters.

[0022] The initial combustion configuration parameters specifically include air supply volume, primary air ratio, secondary air ratio, coal blending ratio, burner start / stop status, and target load status parameters.

[0023] Further, in step S4, the emission efficiency adaptive balancing is used to achieve a strategic trade-off between emission compliance and combustion performance for coal-fired power plants. Specifically, based on the predicted load sequence data, the initial combustion configuration parameters, and multi-source operational data of the coal-fired power plant, an improved multi-constraint rolling weighted balancing method based on dynamic causal control graphs is used to perform emission efficiency adaptive balancing to obtain balancing optimization strategy parameters, including the following steps:

[0024] Step S41: Construct an emission causal control graph to model the dynamic relationship between pollutant emission response and key combustion regulation parameters. Specifically, based on historical operating load data and predicted load scenarios, construct a Bayesian causal graph with nitrogen oxide concentration, sulfur dioxide concentration, carbon dioxide concentration and particulate matter concentration as target nodes. Input nodes include air supply volume, coal type ratio, primary and secondary air ratio and burner status. Through structure learning and edge weight training, obtain emission causal paths and regulation sensitivity scores.

[0025] Step S42: Construct a dynamic evaluation model for emission levels to quantify future emission pressure levels under the background of predicted load trends and initial parameters. Specifically, in combination with the emission causal path, input the predicted load sequence data and initial configuration parameters into the graph model, calculate the estimated concentration sequence of various pollutants in the future time series, calculate the pollutant emission offset factor, evaluate the emission risk level based on weighted rules, and divide it into three levels: "low risk", "medium risk" and "high risk".

[0026] Step S43: Multi-constraint rolling weighted adjustment, used to balance and correct the strategy parameters based on the current emission level and causal feedback path. Specifically, it involves constructing a weighted objective function that includes emission response sensitivity, load regulation stability and current configuration disturbance cost; and adjusting the parameter configuration priority weights in a rolling window manner according to the dynamic feedback window to obtain the optimized strategy parameter set.

[0027] Step S44: Strategy screening and correction, used to ensure that the generated strategy parameters are physically feasible and operable in actual operation. Specifically, it involves constructing a rule constraint model, screening out strategy vectors that violate the upper and lower limits of air volume, the non-mixing ratio of coal, or the minimum number of burners to start and stop, and performing minimum disturbance correction on edge feasible strategies to obtain a balanced optimization strategy parameter set.

[0028] The parameter set of the balance optimization strategy specifically includes a priority vector of adjustment variables, pollutant regulation sensitivity parameters, emission risk level labeling, adjustment weights, and adjustment instructions.

[0029] Furthermore, in step S5, the adaptive optimization of operating parameters is used to adjust and match key operating control parameters of the coal-fired power plant in a responsive manner based on the dynamic load forecast results, so as to achieve the comprehensive goals of load stability, energy efficiency improvement and emission compliance. Specifically, based on the initial combustion configuration parameters and the balance optimization strategy parameters, the predicted load sequence data, environmental emission detection data and historical operation feedback data are integrated to dynamically generate an adaptive control operating parameter set including air supply volume, primary and secondary air ratio, coal blending ratio and burner start-up and shutdown status.

[0030] The present invention provides an adaptive optimization system for operating parameters of a coal-fired power plant, comprising a data fusion module, a load modeling module, a coordination optimization module, an adaptive balancing module, and a parameter optimization module;

[0031] The data fusion module is used to perform data fusion acquisition, obtain multi-source data on the operation of coal-fired power plants through data fusion acquisition, and send the multi-source data on the operation of coal-fired power plants to the load modeling module, the coordination optimization module, and the adaptive balance module.

[0032] The load modeling module is used for dynamic load response modeling. Through dynamic load response modeling, predicted load sequence data is obtained, and the predicted load sequence data is sent to the coordination optimization module and the adaptive balance module.

[0033] The coordination and optimization module is used for coal load coordination and optimization. Through coal load coordination and optimization, the initial combustion configuration parameters are obtained, and the initial combustion configuration parameters are sent to the adaptive balance module and the parameter optimization module.

[0034] The adaptive balancing module is used for emission efficiency adaptive balancing. Through emission efficiency adaptive balancing, it obtains balancing optimization strategy parameters and sends the balancing optimization strategy parameters to the parameter optimization module.

[0035] The parameter optimization module is used for adaptive optimization of operating parameters, and obtains an adaptive control operating parameter set through adaptive optimization of operating parameters.

[0036] The beneficial effects achieved by the present invention using the above solution are as follows:

[0037] (1) In the existing adaptive optimization of operating parameters of coal-fired power plants, there are technical problems such as coarse granularity of multi-source data fusion, low accuracy of load prediction and rigidity of regulation response mechanism. This makes it difficult for the control system to respond effectively in a timely manner when the load fluctuates sharply or the coal quality changes drastically. This solution creatively integrates three functional modules: dynamic load response modeling, coal load coordination optimization and emission efficiency adaptive balance, and constructs a closed-loop adaptive regulation link with multi-module coupling. It can realize accurate fusion processing of multi-dimensional operating data, minute-level prediction of future load and dynamic optimization configuration of control parameters in actual working conditions.

[0038] (2) In view of the technical problems in the existing coal-fired load coordination optimization methods, such as insufficient identification of key adjustment variables and deviation of parameter configuration results from system stability constraints, especially in multi-variable linkage adjustment scenarios, traditional methods cannot take into account the complex coupling relationship between load response, energy efficiency and emission targets. This scheme creatively introduces an adaptive two-stage collaborative scheduling method. First, a coarse-grained configuration is achieved by linear mapping, and then a nonlinear fine adjustment is performed by a multi-objective optimization algorithm. Finally, a stable parameter set is constructed through simulation verification, which effectively improves the feasibility and engineering practicality of parameter generation.

[0039] (3) In view of the technical problem that the existing emission efficiency adaptation balance method has the ability to dynamically model the causal relationship between adjustment parameters and pollutant emissions, which leads to the lag in strategy adjustment and frequent occurrence of pollutant exceedance, this solution creatively introduces a multi-constraint rolling weighted balance method based on dynamic causal control diagram, constructs a causal feedback model with emission target as the guide and combined with sensitivity score, and designs a dynamic priority adjustment mechanism with rolling window, realizing the combination of emission level assessment, adjustment instruction optimization and strategy screening, and enhancing the foresight and compliance assurance capability of strategy adjustment. Attached Figure Description

[0040] Figure 1 A flowchart illustrating an adaptive optimization method for operating parameters of a coal-fired power plant provided by this invention;

[0041] Figure 2 A schematic diagram of an adaptive optimization system for operating parameters of a coal-fired power plant provided by the present invention;

[0042] Figure 3 This is a flowchart illustrating the coal load coordination optimization process in step S3.

[0043] Figure 4 A schematic diagram of the process for adjusting emission efficiency in step S4.

[0044] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0045] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0046] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship 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 limitations on this invention.

[0047] Example 1, see Figure 1 The present invention provides an adaptive optimization method for operating parameters of a coal-fired power plant, the method comprising the following steps:

[0048] Step S1: Run data fusion and acquisition;

[0049] Step S2: Dynamic load response modeling;

[0050] Step S3: Coal-fired load coordination optimization;

[0051] Step S4: Emission efficiency adaptation balance;

[0052] Step S5: Adaptive optimization of runtime parameters.

[0053] By performing the above operations, this solution addresses the technical problems in existing adaptive optimization of operating parameters for coal-fired power plants, such as coarse granularity of multi-source data fusion, low accuracy of load prediction, and rigidity of regulation response mechanisms. These problems make it difficult for the control system to respond effectively in a timely manner when there are drastic load fluctuations or changes in coal quality. This solution creatively integrates three functional modules: dynamic load response modeling, coal-fired load coordination optimization, and emission efficiency adaptive balancing. It constructs a closed-loop adaptive regulation link with multiple modules coupled together, which can achieve accurate fusion processing of multi-dimensional operating data, minute-level prediction of future load, and dynamic optimization configuration of control parameters in actual operating conditions.

[0054] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S1, the operation data fusion and acquisition is used to collect and fuse multi-dimensional data on the operation status of coal-fired power plants under different operating conditions. Specifically, it collects coal-fired power plant operation parameter data through multi-source data acquisition, including coal characteristic data, real-time operating condition data, control system parameters, historical operating load data, and environmental emission detection data.

[0055] By performing unified timestamp alignment, missing value imputation, outlier removal, feature standardization, and category mapping on the operating parameter data of coal-fired power plants, and constructing a standardized operating status vector sequence based on a preset field structure, the fused multi-source operating data of coal-fired power plants is obtained for subsequent predictive modeling and optimization analysis.

[0056] The coal-fired power plant operates using multi-source data, which is in the form of a feature vector sequence with a unified structure.

[0057] Preferably, the multi-source data acquisition specifically includes data acquisition from the boiler and turbine process control system, data acquisition from the coal conveying and pulverizing system, data acquisition from the emission monitoring system, historical database data reading, and manual sampling and analysis data acquisition.

[0058] Table 1 shows a sample table of collected operating parameter data for the coal-fired power plant. As shown in the table, the coal characteristic data specifically includes coal type, lower heating value, volatile matter, ash content, moisture, and sulfur content. Coal type is categorized data, including lean coal, bituminous coal, anthracite, lignite, long-flame coal, weakly caking coal, fat coal, coking coal, gas coal, and lean coal. Lower heating value is a continuous numerical data, expressed in kcal / kg, ranging from 3000 to 6500 kcal / kg. kcal / kg, accurate to one decimal place; volatile matter is a percentage, ranging from 5.0% to 45.0%, with an accuracy of 0.1%; ash content is a percentage, ranging from 5.0% to 40.0%, with an accuracy of 0.1%; moisture content is a percentage, ranging from 0.0% to 25.0%, with an accuracy of 0.1%; sulfur content is a percentage, ranging from 0.00% to 5.00%, with an accuracy of 0.01%.

[0059] The real-time operating data specifically includes unit load, main steam temperature, main steam pressure, furnace temperature, oxygen content, and flue gas temperature. The unit load is continuous data in megawatts (MW), ranging from 100 MW to 660 MW, suitable for conventional coal-fired power generation units. The main steam temperature is in degrees Celsius (°C), ranging from 450°C to 570°C with an accuracy of 1°C. The main steam pressure is in megapascals (MPa), ranging from 9 MPa to 16 MPa with an accuracy of 0.1 MPa. The furnace temperature is in degrees Celsius (°C), ranging from 1000°C to 1600°C with an accuracy of 1°C. The oxygen content is a percentage, limited to 2.00% to 8.00% with an accuracy of 0.01%. The flue gas temperature is in degrees Celsius (°C), ranging from 100°C to 250°C with an accuracy of 1°C.

[0060] The control system parameters specifically include burner start / stop status, primary air to secondary air ratio, air supply volume, and coal blending strategy. The burner start / stop status is Boolean or enumerated data, ranging from "on" to "off," supporting up to 16 burner channels numbered A1 to A16. The primary air to secondary air ratio is a ratio data representing the air volume proportion, using standardized floating-point form, with values ​​satisfying a sum of 1. The air supply volume is a continuous numerical data, in standard cubic meters per hour (Nm³ / h), ranging from 20,000 Nm³ / h to 100,000 Nm³ / h, with an accuracy of 100 Nm³ / h. The coal blending strategy is a proportional vector data describing the mixing ratio of two or three coal types, supporting three-coal blending, with a sum of proportions of 100% and an accuracy of 1%.

[0061] The historical operating load data specifically includes historical load change sequences, oxygen concentration change sequences, and main steam temperature change sequences. All three fields are time series data, with a sampling interval of 1 minute and a sampling time window limited to the previous 60 to 120 minutes. Each sequence contains a sequence vector consisting of 60 to 120 floating-point values. The unit of the historical load sequence is megawatts (MW), the unit of the oxygen concentration sequence is percentage (%), and the unit of the main steam temperature sequence is degrees Celsius (°C). All of these correspond to continuous data windows up to the current moment, forming a dynamic operating trend within a continuous time period.

[0062] The environmental emission monitoring data specifically includes carbon dioxide concentration, nitrogen oxide emission concentration, sulfur dioxide emission concentration, and particulate matter concentration. Carbon dioxide concentration is presented as a percentage, ranging from 10.00% to 15.00%, with an accuracy of 0.01%. Nitrogen oxide (NOx) emission concentration is presented as continuous data, measured in milligrams per cubic meter (mg / m³), ranging from 20 mg / m³ to 400 mg / m³, with an accuracy of 1 mg / m³. Sulfur dioxide (SO2) emission concentration is presented as continuous data, measured in mg / m³, ranging from 10 mg / m³ to 800 mg / m³, with an accuracy of 1 mg / m³. Particulate matter concentration is presented as continuous data, measured in mg / m³, ranging from 0 mg / m³ to 100 mg / m³, with an accuracy of 1 mg / m³.

[0063] Table 1. Sample Data Collection of Operating Parameters from Coal-fired Power Plants

[0064]

[0065] Example 3, see Figure 1 , Figure 2 This embodiment is based on the above embodiment. In step S2, the dynamic load response modeling is used to construct a load response prediction model for coal-fired power plants under different operating conditions and control strategies, thereby improving the foresight and scheduling adaptability of load regulation. Specifically, based on the multi-source data of the coal-fired power plant operation, a standard long short-term memory network is used to perform time modeling on the historical load sequence. Combined with the current operating condition characteristics as input, a rolling prediction of the load trend for the next 60 minutes is achieved, resulting in predicted load sequence data.

[0066] The predicted load sequence data is used to represent the future load trend prediction results output by the model, in megawatts (MW), and corresponds to the minute-by-minute prediction values ​​for the next 60 minutes;

[0067] Preferably, Table 2 is an example table of parameter settings for the standard long short-term memory network. As shown in the table, the input dimension of the standard long short-term memory network is set to 18, which is aligned with the feature dimension in the multi-source data of the coal-fired power plant operation. The loss function adopts the mean squared error loss function, and two layers of long short-term memory are superimposed to enhance the temporal modeling.

[0068] Table 2. Example of parameter settings for a standard Long Short-Term Memory (LSTM) network.

[0069]

[0070] Preferably, the input dimension of the standard Long Short-Term Memory network adopts an 18-dimensional feature design scheme to characterize the power plant's operating state at each time step, specifically including the following three categories:

[0071] The time-series operating condition characteristics (9 dimensions) include unit load, oxygen content, main steam temperature, main steam pressure, furnace temperature, flue gas temperature, air supply volume, primary air ratio, and load change rate. Among them, the primary air ratio and secondary air ratio satisfy the constraint that their sum is 1 to avoid input dimension redundancy. The load change rate is used to characterize the dynamic change trend of the load over time, thereby improving the model's ability to perceive and predict rapid ramp-up or sudden drop conditions.

[0072] Coal quality and blending characteristics (5 dimensions), including lower heating value, volatile matter, ash content, moisture and sulfur content, are used to reflect the quality differences and combustion characteristics of current coal, thereby improving the robustness of the model to different fuels.

[0073] The control and structural summary features (4 dimensions) include the number of burners in operation, the start-stop imbalance between upper and lower layers of burners, the coal blending ratio degree of freedom 1, and the coal blending ratio degree of freedom 2. Among them, the number of burners in operation is used to characterize the total number of burners that start and stop, and the upper and lower layer imbalance is used to characterize the differences in the spatial distribution of burners. The coal blending ratio degree of freedom can uniquely determine the three-coal blending ratio by constraining the two degrees of freedom, thereby achieving an effective characterization of the coal blending structure while avoiding redundant input.

[0074] Through the above design, an 18-dimensional feature vector is constructed at each time step. Combined with the time window of the previous 60 minutes, an input sequence is formed and input into the long short-term memory network model for modeling, so as to realize the load prediction for the next 60 minutes minute by minute.

[0075] Example 4, see Figure 1 , Figure 2 and Figure 3This embodiment is based on the above embodiment. In step S3, the coal-fired load coordination optimization, based on the predicted load trend, performs coordinated optimization configuration among various operating sub-parameters in the coal-fired power generation system to achieve a dynamic balance between combustion efficiency, load response speed, and emission control. Specifically, based on the predicted load sequence data and multi-source operating data of the coal-fired power plant, an adaptive two-stage coordinated scheduling method is used to perform coal-fired load coordination optimization to obtain the initial combustion configuration parameters, including the following steps:

[0076] Step S31: Identification of key load regulation variables, used to identify the regulation variables most sensitive to predicted load changes from multi-source input features. Specifically, based on coal characteristic data, control system parameters and predicted load sequence data, a feature importance assessment model is constructed. An improved Shapley value method and cross-information gain fusion strategy are used to extract a set of key regulation parameters that have a high response to load changes, including air supply volume, primary and secondary air ratio, coal blending ratio and burner channel status.

[0077] Preferably, the feature importance evaluation model adopts a gradient boosting decision tree (GBDT) model based on ensemble learning. By gradually fitting the residuals in multiple iterations, a nonlinear mapping relationship between the predicted load change and the input feature variables is established. This model has the ability to handle high-dimensional nonlinear features and capture the interaction effects between variables. After training, it can output the marginal contribution value of each feature variable, providing a basis for Shapley value calculation.

[0078] Preferably, the improved Shapley value method and cross-information gain fusion strategy are used to comprehensively calculate the importance of each variable in the coal combustion characteristic data, control system parameters, and predicted load sequence data, thereby achieving feature importance assessment. The calculation formula is as follows:

[0079] ;

[0080] In the formula, This is the importance score of the j-th variable, where j is the index of the feature variable in the coal combustion characteristic data, control system parameters, and predicted load sequence data. This is the feature importance weight, with a default value of 0.6, according to Shapley. j It is the original Shapley value of feature contribution output by the feature importance assessment model, InfoGain. j It is the mutual information gain value between the j-th feature variable and the future load change, which is specifically estimated based on the conditional probability distribution between the predicted load sequence data and the feature variable;

[0081] Step S32: Initial adjustment parameter generation, used to construct the first stage coarse-grained adjustment parameter configuration while maintaining the tracking of the predicted load. Specifically, it constructs a linear regression mapping relationship based on constrained least squares to map the predicted load sequence to the adjustment variable space, and introduces the historical operating load data to form upper and lower limit constraints to obtain the initial vector of combustion initial settings, including the initial air supply volume, primary and secondary air ratio and coal type ratio vector.

[0082] Step S33: Multi-objective collaborative refinement optimization, used to perform nonlinear multi-objective refinement adjustment for the three major objectives of energy saving, environmental protection and response rate. Specifically, it constructs a multi-objective optimization algorithm based on improved differential evolution, with the objective functions of maximizing combustion efficiency, minimizing unit load emissions and minimizing response delay. It integrates the predicted load gradient and operating dynamic conditions to iteratively optimize the initial parameters and generate the optimal configuration parameters that meet the objective constraints.

[0083] Preferably, the objective function is calculated using the following formula:

[0084] ;

[0085] In the formula, F is the objective function, u is a vector in the initial vector of combustion initial settings, used as an index of the configuration parameters to be optimized, f1 is the heat consumption target, the objective is to minimize it to represent the maximum combustion efficiency, f2 is the emission target per unit load, the objective is to minimize it, and f3 is the load response delay target, the objective is to minimize it.

[0086] Preferably, the formula for calculating the heat consumption target is:

[0087] ;

[0088] In the formula, LHV(c) is the coal consumption per unit time, LHV(c) is the lower heating value under the coal blending ratio, c is the coal blending ratio parameter, and P(u) is the unit output power.

[0089] Preferably, the formula for calculating the unit load emission target is as follows;

[0090] ;

[0091] In the formula, It is an emissions index, specifically including values ​​for nitrogen oxides, sulfur dioxide, carbon dioxide, and particulate matter. It is the emission weighting factor, which is manually set based on the power plant's operation. E k (u) represents emissions;

[0092] Preferably, the calculation formula for the load response delay target is:

[0093] ;

[0094] In the formula, T resp (u) is the response time required for the load change to reach the target. This is the perturbation cost weighting coefficient, with a value range of [0.01, 0.2], and a default value of 0.05. prev These are the configuration parameters from the previous moment; ||·||2 is the L2 norm operator.

[0095] Step S34: Parameter stability optimization, used to dynamically simulate and verify the stability of the output parameters and the feasibility of the system under actual working conditions. Specifically, the optimized configuration parameters are input into the simulation module based on physical model constraints to perform thermal balance, air-coal flow field simulation and emission prediction. If the output indicators meet the safety boundary and response requirements, the configuration is confirmed as the initial combustion configuration parameters; otherwise, the process returns to step S33 for reconstruction iteration to obtain the initial combustion configuration parameters.

[0096] Preferably, the simulation module based on physical model constraints performs physical constraints and simulation modeling based on simplified thermal equilibrium relationships, and the calculation formula is as follows:

[0097] ;

[0098] In the formula, Q in Q is the heat input from coal combustion. steam Q is the effective heat absorbed by the steam. loss It is the system that loses heat;

[0099] The initial combustion configuration parameters specifically include air supply volume, primary air ratio, secondary air ratio, coal blending ratio, burner start / stop status, and target load status parameters.

[0100] By performing the above operations, this solution addresses the technical problems in existing coal-fired load coordination and optimization methods, such as insufficient identification of key adjustment variables and deviation of parameter configuration results from system stability constraints. In particular, in multi-variable linkage adjustment scenarios, traditional methods cannot take into account the complex coupling relationship between load response, energy efficiency, and emission targets. This solution creatively introduces an adaptive two-stage collaborative scheduling method. First, a coarse-grained configuration is achieved using linear mapping, and then a nonlinear refinement adjustment is performed using a multi-objective optimization algorithm. Finally, a stable parameter set is constructed through simulation verification, which effectively improves the feasibility and engineering practicality of parameter generation.

[0101] Example 5, see Figure 1 , Figure 2 and Figure 4This embodiment is based on the above embodiment. In step S4, the emission efficiency adaptation balance is used to realize the strategic trade-off between emission compliance and combustion performance of coal-fired power plants. Specifically, based on the predicted load sequence data, the initial combustion configuration parameters, and multi-source operation data of the coal-fired power plant, an improved multi-constraint rolling weighted balance method based on dynamic causal control graph is used to perform emission efficiency adaptation balance to obtain balance optimization strategy parameters, including the following steps:

[0102] Step S41: Construct an emission causal control graph to model the dynamic relationship between pollutant emission response and key combustion regulation parameters. Specifically, based on historical operating load data and predicted load scenarios, construct a Bayesian causal graph with nitrogen oxide concentration, sulfur dioxide concentration, carbon dioxide concentration, and particulate matter concentration as target nodes. Input nodes include air supply volume, coal type ratio, primary and secondary air ratio, and burner status. Through structure learning and edge weight training, the emission causal control graph is obtained. The emission causal control graph specifically includes emission causal paths and regulation sensitivity scores. The emission causal paths are used to represent the graph structure of the emission causal control graph, that is, to describe the causal connection relationship between input nodes and target nodes. The regulation sensitivity scores are used to represent the edge weights of the emission causal control graph, that is, to characterize the strength and sensitivity of the causal connection relationship.

[0103] Preferably, the structure learning employs a causal structure learning method based on joint likelihood and sparsity constraints. Causal structure learning is performed by constructing a structure learning objective function, the formula for which is:

[0104] ;

[0105] In the formula, F G Here, G is the structural learning objective function, G is a Bayesian causal graph with nitrogen oxide concentration, sulfur dioxide concentration, carbon dioxide concentration, and soot concentration as objective nodes, and D is historical operating load data. It is the negative log-likelihood calculation function. is the sparsity constraint coefficient, with a value range of [0.01, 0.2], W(G) is the set of edge weights, and ||·||1 is the L1 norm operator;

[0106] Step S42: Construct a dynamic evaluation model for emission levels to quantify future emission pressure levels under the background of predicted load trends and initial parameters. Specifically, in conjunction with the emission causal path in step S41, input the predicted load sequence and initial configuration parameters into the graph model, calculate the estimated concentration sequence of various pollutants in the future time series, calculate the pollutant emission offset factor, and evaluate the emission risk level based on weighted rules, classifying it into three levels: "low risk", "medium risk" and "high risk".

[0107] Preferably, the formula for calculating the pollutant emission offset factor is:

[0108] ;

[0109] In the formula, It is the pollutant emission offset factor, where k is the emission index. It is a prediction of emission concentrations. These are the pollutant emission limits, which are manually set based on industry-specific regulations; t is the time index.

[0110] Step S43: Multi-constraint rolling weighted adjustment, used to balance and correct the strategy parameters based on the current emission level and causal feedback path. Specifically, it involves constructing a weighted objective function that includes emission response sensitivity, load regulation stability and current configuration disturbance cost; and adjusting the parameter configuration priority weights in a rolling window manner according to the dynamic feedback window to obtain the optimized strategy parameter set.

[0111] Preferably, the formula for calculating the weighted objective function is:

[0112] ;

[0113] In the formula, J t It is a weighted objective function, where w1(t) is the emission response sensitivity weight. It is the emission response sensitivity, specifically obtained by weighting the regulation sensitivity score in step S41, where w2(t) is the load regulation stability weight, and Stab... t This is a load regulation stability parameter, w3(t) is the current configuration disturbance cost weight, and Cost is... t This is the cost of the current configuration disturbance;

[0114] The formula for calculating the priority weight of parameter configuration based on the dynamic feedback window and using a scrolling window is as follows:

[0115] ;

[0116] In the formula, w t It is the set of weights in the weighted objective function adjusted using a rolling window method. This is the smoothing coefficient, with a value range of [0.6, 0.9], and a default value of 0.6. It is the normalized emission response sensitivity, h(Stab) t ) is the normalized load regulation stability parameter, c(Cost) t This is the normalized current configuration perturbation cost;

[0117] Step S44: Strategy screening and correction, used to ensure that the generated strategy parameters are physically feasible and operable in actual operation. Specifically, a rule constraint model is constructed to screen out strategy vectors that violate the upper and lower limits of air volume, the non-mixing ratio of coal, or the minimum number of burners to start and stop, and the marginal feasibility strategies are corrected by minimum disturbance to obtain a balanced optimization strategy parameter set.

[0118] Preferably, the calculation formula for the rule constraint model is:

[0119] ;

[0120] In the formula, It is a set of parameters for a balanced optimization strategy. The overall constraint is the minimum projective perturbation constraint. It is the set of feasible strategy parameters, used to represent the strategy index in the set of all strategy parameters that satisfy the constraints. This is a rule constraint domain used to represent upper and lower limits of air volume, non-mixing ratio of coal, and minimum number of burners to start and stop. It is a candidate policy parameter vector, used to represent the set of optimization policy parameters, when satisfying When the corresponding optimization strategy is selected, it is retained; otherwise, it is discarded. This is the constraint filtering correction threshold, specifically set at 20% of the range of the variables in each rule's constraint domain;

[0121] The parameter set of the balance optimization strategy specifically includes a priority vector of adjustment variables, pollutant regulation sensitivity parameters, emission risk level labeling, adjustment weights, and adjustment instructions.

[0122] By performing the above operations, this solution addresses the technical problem in existing emission efficiency adaptation balancing methods where the dynamic modeling capability of the causal relationship between adjustment parameters and pollutant emissions leads to lags in strategy adjustments and frequent exceedances of pollutant standards. It creatively introduces a multi-constraint rolling weighted balancing method based on a dynamic causal control diagram, constructs a causal feedback model guided by emission targets and incorporating sensitivity scoring, and designs a dynamic priority adjustment mechanism with a rolling window. This achieves a combination of emission level assessment, adjustment command optimization, and strategy selection, enhancing the foresight and compliance assurance capabilities of strategy adjustments.

[0123] Example 6, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S5, the adaptive optimization of operating parameters is used to adjust and match the key operating control parameters of the coal-fired power plant in a responsive manner based on the dynamic load forecast results, so as to achieve the comprehensive goal of load stability, energy efficiency improvement and emission compliance. Specifically, based on the initial combustion configuration parameters and the balance optimization strategy parameters, the predicted load sequence data, environmental emission detection data and historical operation feedback data are integrated to dynamically generate an adaptive control operating parameter set including air supply volume, primary and secondary air ratio, coal blending ratio and burner start-up and shutdown status.

[0124] Example 7, see Figure 1 and Figure 2 Based on the above embodiments, this embodiment provides an adaptive optimization system for operating parameters of a coal-fired power plant, including a data fusion module, a load modeling module, a coordination optimization module, an adaptive balancing module, and a parameter optimization module.

[0125] The data fusion module is used to perform data fusion acquisition, obtain multi-source data on the operation of coal-fired power plants through data fusion acquisition, and send the multi-source data on the operation of coal-fired power plants to the load modeling module, the coordination optimization module, and the adaptive balance module.

[0126] The load modeling module is used for dynamic load response modeling. Through dynamic load response modeling, predicted load sequence data is obtained, and the predicted load sequence data is sent to the coordination optimization module and the adaptive balance module.

[0127] The coordination and optimization module is used for coal load coordination and optimization. Through coal load coordination and optimization, the initial combustion configuration parameters are obtained, and the initial combustion configuration parameters are sent to the adaptive balance module and the parameter optimization module.

[0128] The adaptive balancing module is used for emission efficiency adaptive balancing. Through emission efficiency adaptive balancing, it obtains balancing optimization strategy parameters and sends the balancing optimization strategy parameters to the parameter optimization module.

[0129] The parameter optimization module is used for adaptive optimization of operating parameters, and obtains an adaptive control operating parameter set through adaptive optimization of operating parameters.

[0130] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0131] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0132] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An adaptive optimization method for operating parameters of a coal-fired power plant, characterized in that: The method includes the following steps: Step S1: Data fusion and acquisition: Collect operating parameter data of coal-fired power plants and fuse them to obtain multi-source operating data of coal-fired power plants; Step S2: Dynamic load response modeling. Based on the multi-source data of the coal-fired power plant operation, perform time modeling on the historical load sequence to obtain the predicted load sequence data. Step S3: Coal-fired load coordination optimization. Based on the predicted load sequence data and multi-source data of coal-fired power plant operation, an adaptive two-stage collaborative scheduling method is adopted to optimize the coal-fired load coordination and obtain the initial configuration parameters for combustion. This includes the following steps: identification of key variables for load regulation, generation of preliminary regulation parameters, multi-objective collaborative refinement optimization, and parameter stability optimization. The multi-objective collaborative refinement optimization constructs a multi-objective optimization algorithm based on improved differential evolution, with the objective functions being maximum combustion efficiency, minimum unit load emissions, and minimum response delay. It integrates predicted load gradients and dynamic operating conditions to iteratively optimize the initial parameters and obtain the optimal configuration parameters. The parameter stability optimization involves inputting the optimal configuration parameters into a simulation module based on physical model constraints to perform thermal balance, air-coal flow field simulation, and emission prediction. If the output indicators meet the safety boundary and response requirements, the configuration is confirmed as the initial combustion configuration parameters; otherwise, the multi-objective collaborative refinement optimization steps are repeated for reconstruction iteration to obtain the initial combustion configuration parameters. Step S4: Emission efficiency adaptation balance. Based on the predicted load sequence data, the initial combustion configuration parameters, and multi-source operation data of the coal-fired power plant, an improved multi-constraint rolling weighted balance method based on dynamic causal control diagram is used to perform emission efficiency adaptation balance and obtain balance optimization strategy parameters. This includes the following steps: constructing emission causal control diagram, constructing emission level dynamic evaluation model, multi-constraint rolling weighted adjustment, and strategy screening and correction. The multi-constraint rolling weighted adjustment constructs a weighted objective function that includes emission response sensitivity, load adjustment stability, and the current configuration disturbance cost, thereby obtaining an optimization strategy parameter set; The strategy screening and correction involves constructing a rule-constrained model, filtering out strategy vectors that violate the constraints, and obtaining a balanced optimization strategy parameter set. The balanced optimization strategy parameter set specifically includes a priority vector of adjustment variables, pollutant regulation sensitivity parameters, emission risk level labels, adjustment weights, and adjustment instructions. Step S5: Adaptive optimization of operating parameters. Based on the initial combustion configuration parameters and the balance optimization strategy parameters, the predicted load sequence data, environmental emission detection data and historical operating load data are integrated to dynamically generate an adaptive control operating parameter set.

2. The adaptive optimization method for operating parameters of a coal-fired power plant according to claim 1, characterized in that: In step S1, the operating parameter data of the coal-fired power plant includes coal characteristic data, real-time operating condition data, control system parameters, historical operating load data, and environmental emission monitoring data.

3. The adaptive optimization method for operating parameters of a coal-fired power plant according to claim 2, characterized in that: In step S3, the identification of key variables for load regulation involves constructing a feature importance assessment model based on coal combustion characteristic data, control system parameters, and predicted load sequence data, and using an improved Shapley value method and a cross-information gain fusion strategy to obtain a set of key regulation parameters. The initial adjustment parameters are generated by constructing a linear regression mapping relationship based on constrained least squares, mapping the predicted load sequence to the adjustment variable space, and introducing the historical operating load data to form upper and lower limit constraints, thereby obtaining the initial vector of the combustion initial setting value.

4. The adaptive optimization method for operating parameters of a coal-fired power plant according to claim 3, characterized in that: In step S3, the initial combustion configuration parameters specifically include air supply volume, primary air ratio, secondary air ratio, coal blending ratio, burner start / stop status, and target load status parameters.

5. The adaptive optimization method for operating parameters of a coal-fired power plant according to claim 4, characterized in that: In step S4, an emission causal control graph is constructed, a Bayesian causal graph is constructed, and emission causal paths and control sensitivity scores are obtained through structure learning and edge weight training. The proposed emission level dynamic evaluation model, combined with the emission causal path, calculates the estimated concentration sequence of pollutants and the pollutant emission offset factor for future time series, and weights the emission risk level for assessment.

6. An adaptive optimization system for operating parameters of a coal-fired power plant, used to implement the adaptive optimization method for operating parameters of a coal-fired power plant as described in any one of claims 1-5, characterized in that: It includes a data fusion module, a load modeling module, a coordination optimization module, an adaptive balancing module, and a parameter optimization module; The data fusion module is used to perform data fusion acquisition, obtain multi-source data on the operation of coal-fired power plants through data fusion acquisition, and send the multi-source data on the operation of coal-fired power plants to the load modeling module, the coordination optimization module, and the adaptive balance module. The load modeling module is used for dynamic load response modeling. Through dynamic load response modeling, predicted load sequence data is obtained, and the predicted load sequence data is sent to the coordination optimization module and the adaptive balance module. The coordination and optimization module is used for coal load coordination and optimization. Through coal load coordination and optimization, the initial combustion configuration parameters are obtained, and the initial combustion configuration parameters are sent to the adaptive balance module and the parameter optimization module. The adaptive balancing module is used for emission efficiency adaptive balancing. Through emission efficiency adaptive balancing, it obtains balancing optimization strategy parameters and sends the balancing optimization strategy parameters to the parameter optimization module. The parameter optimization module is used for adaptive optimization of operating parameters, and obtains an adaptive control operating parameter set through adaptive optimization of operating parameters.

Images