Optimization method for improving flexible operation of thermal power unit

Abstract

The present invention relates to the technical field of big data, and provides an optimization method for improving flexible operation of a thermal power unit. The method comprises: analyzing first operating state data obtained by monitoring the operating state of a target unit in real time, so as to obtain a unit operating state result; using a parameter prediction model constructed by means of a machine learning algorithm to predict key control parameters of the target unit on the basis of the first operating state data, so as to obtain a control parameter prediction result; and analyzing the control parameter prediction result and the unit operating state result to generate a target regulation strategy to act on the target unit. The present invention monitors in real time and analyzes the operating state of the unit to acquire the operating state of the unit at the current moment, and takes into account and analyzes the control parameter prediction result of the unit predicted by using the parameter prediction model constructed by means of the machine learning algorithm, so as to generate the target regulation strategy to act on the unit, thereby maximizing unit benefits while implementing intelligent control over key control loops of units and regulating parameters flexibility.

Description

An optimization method to improve the flexible operation of thermal power units Technical Field

[0001] This invention relates to the field of big data technology, and in particular to an optimization method for improving the flexible operation of thermal power units. Background Technology

[0002] Thermal power units have complex production processes, and their equipment systems operate in environments characterized by high temperature, high pressure, high dust, and high corrosion for extended periods. As the units operate continuously for long periods, the equipment gradually ages, and its characteristics change significantly, leading to a decline in the performance of the unit's control system. This forces operators to frequently intervene manually, greatly increasing the workload and monitoring pressure. Currently, the operation of most units still relies heavily on the experience of operators, and from the perspective of safe operation, it is difficult to make flexible adjustments to adapt to changes in the unit, while also neglecting the unit's economic efficiency. Therefore, how to achieve flexible adjustments to the unit's key control loops while improving economic efficiency has become one of the current research focuses.

[0003] Therefore, the present invention provides an optimized method for improving the flexible operation of thermal power units. Summary of the Invention

[0004] This invention provides an optimization method to improve the flexible operation of thermal power units. By monitoring and analyzing the operating status of the unit in real time, the current operating status of the unit is obtained, and the analysis is combined with the predicted control parameters of the unit obtained by the parameter prediction model built using machine learning algorithm to generate a target adjustment strategy to be applied to the unit. This achieves intelligent control of the key control loops of the unit, flexible adjustment of parameters, and maximizes the unit's benefits at the same time.

[0005] This invention provides an optimization method for improving the flexible operation of thermal power units, comprising:

[0006] Step 1: Monitor the operating status of the target unit in real time to obtain the first operating status data;

[0007] Step 2: Based on the first operating status data, analyze the current operating status of the target unit to obtain the unit operating status result;

[0008] Step 3: Using a parameter prediction model constructed with machine learning algorithms, predict the key control parameters of the target unit based on the first operating status data to obtain the control parameter prediction results;

[0009] Step 4: Combine the predicted control parameters with the unit operating status results for analysis, generate a target adjustment strategy and apply it to the target unit to maximize unit efficiency.

[0010] Preferably, the first operating status data includes power generation, efficiency, pressure and temperature, operating load, fuel consumption and emissions, unit flow rate, and operating time.

[0011] Preferably, based on the first operating status data, the operating status of the current target unit is analyzed to obtain the unit operating status result, including:

[0012] Obtain the operating data threshold table for each operating system within the target unit;

[0013] Based on the running data threshold table, the first running status data of each running system is compared with the corresponding running data threshold to obtain the first running difference;

[0014] The first operational difference is compared and analyzed with a preset difference threshold to determine the level labeling result of the first operational status data;

[0015] Based on the level labeling results of the first operating status data of each operating system, calculate the system risk coefficient of the current operating system;

[0016] The formula for calculating the system risk coefficient is as follows:

[0017] In the formula, X represents the system risk coefficient of the currently running system; This represents the number of running state data at the i-th state level within the currently running system, where... ; This represents the total number of running status data points of the currently running system; It is represented as the first operational difference between the j-th operational state data at the i-th state level in the current operating system and the corresponding operational data threshold, where, ; This represents the risk impact weight of the j-th operating state data at the i-th state level within the current operating system;

[0018] The system risk coefficients of each operating system of the target unit are set to generate the unit's operating status results.

[0019] Preferably, a parameter prediction model constructed using machine learning algorithms is used to predict key control parameters of the target unit based on the first operating state data, resulting in control parameter prediction results, including:

[0020] Historical operating status data, equipment parameters, and environmental condition data of the target unit are extracted and preprocessed to obtain the first dataset.

[0021] The first dataset is divided according to a set ratio to obtain a training dataset and a test dataset;

[0022] Extract the correlation and influence features of the key control parameters of the target unit, and use the training dataset as input to train the neural network to obtain an initial prediction model;

[0023] The initial prediction model was evaluated using the test dataset, and the model was tuned based on the evaluation results to obtain the parameter prediction model.

[0024] The first operating state data is input into the parameter prediction model to obtain the control parameter prediction results.

[0025] Preferably, the predicted control parameters are analyzed in conjunction with the unit operating status results to generate a target adjustment strategy that is applied to the target unit to maximize unit efficiency, including:

[0026] Based on the system risk coefficient in the unit operation status results, determine the safety risk level of each operating system at the current moment;

[0027] From the risk level-adjustment strategy database, extract the first state data of each operating system at the current moment - adjustment strategy reference table based on the security risk level;

[0028] From the first state data-adjustment strategy reference table, determine the data adjustment strategy for the first operating state data labeled at different levels within the corresponding operating system, and output it as a reference adjustment strategy;

[0029] Determine the control input parameters for the key control parameters controlling the target unit;

[0030] Parameters that are directly related to the unit's economic performance are considered as associated input parameters, while parameters that are not directly related to the unit's economic performance are considered as non-associated input parameters.

[0031] Based on the reference adjustment strategy, obtain the corresponding parameter adjustment strategy for each key control parameter and the comparison parameter for the non-correlated input parameter;

[0032] Based on the predicted control parameters, the target prediction results of the associated and non-associated input parameters of the key control parameters are obtained by using a target optimization algorithm.

[0033] By combining the adjustment strategies of the associated and non-associated input parameters of each key control parameter with the target prediction results, a target adjustment strategy is generated.

[0034] Preferably, based on the control parameter prediction results, the target prediction results of the associated and non-associated input parameters of the key control parameters are obtained by using a target optimization algorithm, including:

[0035] Obtain the technical and economic performance indicators of key control parameters;

[0036] By taking the associated and non-associated input parameters of key control parameters as input variables and the comprehensive cost of the control loop as the output variable, and combining the corresponding technical performance indicators and economic performance indicators, a control parameter-economic model is established.

[0037] Based on the aforementioned control parameter-economic model, the objective function is to minimize the unit's hourly comprehensive cost.

[0038] The input variables are used as linear constraints, and a parameter optimization prediction model is constructed using an objective optimization algorithm.

[0039] Solving the parameter optimization prediction model yields the first prediction result sequence of the associated and non-associated input parameters of the current key control parameters within a set prediction time period.

[0040] Extract the prediction result at the current sampling time from the first prediction result sequence, and output it as the target prediction result for both associated and unassociated input parameters.

[0041] Preferably, the target adjustment strategy is generated by combining the adjustment strategies of the associated and non-associated input parameters of each key control parameter with the target prediction results, including:

[0042] Based on the combined analysis of the parameter adjustment strategy and the level labeling results of the associated input parameters, the adjustment range of the first parameter is obtained;

[0043] Based on the combined analysis of the parameter adjustment strategy for non-associated input parameters and the results of level labeling, the adjustment range of the second parameter is obtained;

[0044] When the correlation-target prediction results of all associated input parameters of the key control parameter are within the adjustment range of the corresponding first parameter, the correlation-target prediction results are output as the optimized values ​​of the associated input parameters of the corresponding associated input parameters.

[0045] When the target prediction result of one or more associated input parameters of a key control parameter does not fall within the adjustment range of the first parameter, the corresponding associated input parameter is marked as the first parameter.

[0046] The linear constraint condition of the first parameter is adjusted using the constraint adjustment coefficient, and the parameter optimization prediction model is re-solved based on the adjusted linear constraint condition to obtain the target prediction results for associated and non-associated input parameters.

[0047] Analyze the target prediction results of the associated input parameters with the corresponding adjustment range of the first parameter, determine the optimized value of the associated input parameters, and output it.

[0048] If the non-associative target prediction result of the non-associative input parameter falls within the adjustment range of the second parameter, then the current non-associative target prediction result is used as the optimized value of the non-associative input parameter.

[0049] If the non-associative target prediction result of the current non-associative input parameter does not fall within the adjustment range of the second parameter, then calculate the absolute difference between the non-associative target prediction result and the upper limit and lower limit of the second parameter adjustment respectively, and obtain the first absolute difference and the second absolute difference accordingly;

[0050] If the first absolute difference is greater than the second absolute difference, then the upper limit of the second parameter adjustment is used as the optimization value of the non-associative input parameter;

[0051] Otherwise, adjust the lower limit of the second parameter as the optimized value for the non-associative input parameter;

[0052] Based on the optimized values ​​of the associated input parameters and the optimized values ​​of the non-associated input parameters for each key control parameter, a target adjustment strategy is generated.

[0053] Preferably, the formula for calculating the constraint adjustment factor is as follows:

[0054] In the formula, P represents the constraint optimization coefficient; Represented as a linear constraint adjustment factor; This represents the upper limit of the adjustment for the corresponding first parameter; This represents the lower limit of the adjustment of the corresponding first parameter; It is represented as the absolute difference between the current first parameter's correlation-target prediction result and the corresponding first parameter's adjustment upper limit; It is represented as the absolute difference between the current first parameter's correlation-target prediction result and the corresponding first parameter's lower limit; This represents the weight of the current first parameter's impact on unit safety. This represents the contribution compensation coefficient of the current first parameter to the unit's economic efficiency.

[0055] Compared with the prior art, the beneficial effects of this application are as follows:

[0056] By monitoring and analyzing the unit's operating status in real time, the current operating status of the unit is obtained, and combined with the predicted control parameters of the unit obtained by the parameter prediction model built using machine learning algorithms, a target adjustment strategy is generated and applied to the unit. This achieves intelligent control of the unit's key control loops, flexible parameter adjustment, and maximizes the unit's efficiency.

[0057] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0058] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0059] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0060] Figure 1 is a flowchart of an optimization method for improving the flexible operation of thermal power units according to an embodiment of the present invention. Detailed Implementation

[0061] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0062] This invention provides an optimization method for improving the flexible operation of thermal power units, as shown in Figure 1, including:

[0063] Step 1: Monitor the operating status of the target unit in real time to obtain the first operating status data;

[0064] Step 2: Based on the first operating status data, analyze the current operating status of the target unit to obtain the unit operating status result;

[0065] Step 3: Using a parameter prediction model constructed with machine learning algorithms, predict the key control parameters of the target unit based on the first operating status data to obtain the control parameter prediction results;

[0066] Step 4: Combine the predicted control parameters with the unit operating status results for analysis, generate a target adjustment strategy and apply it to the target unit to maximize unit efficiency.

[0067] In this embodiment, the target unit refers to the complete set of equipment and systems that use thermal power to generate electricity; the first operating status data includes power generation, efficiency, pressure and temperature, operating load, fuel consumption and emissions, unit flow rate and operating time; the unit operating status result refers to the system risk coefficient of each operating system of the target unit, wherein the operating system includes the boiler system, turbine system, generator system, auxiliary system and control system; the system risk coefficient is used to characterize the degree of abnormality in the operating status of the current target unit's operating system.

[0068] In this embodiment, the machine learning algorithm refers to an algorithm that automatically learns patterns and rules from data by analyzing and learning data, and uses these patterns and rules to perform tasks such as prediction, classification, clustering, and dimensionality reduction. The parameter prediction model is a model obtained by training a neural network using preprocessed historical operating status data of the unit, equipment parameters, and environmental conditions. It is used to predict the future values ​​of the unit's key control parameters. The key control parameters include main steam temperature, pressure, intermediate point temperature, denitrification efficiency, etc., which refer to the parameters of key control loops such as main steam temperature, pressure, intermediate point temperature, and denitrification.

[0069] In this embodiment, the control parameter prediction result refers to the predicted parameter value obtained by using the parameter prediction model to predict the key control parameters of the target unit based on the first operating state data; the target adjustment strategy is used to adjust the associated input parameters that are directly related to the economy of the key control parameters, as well as the non-associated input parameters that are not directly related to the economy, so as to achieve precise control of the key control parameters and ensure the flexibility of the target unit and the maximization of the unit's benefits.

[0070] The beneficial effects of the above technical solution are: by monitoring and analyzing the operating status of the unit in real time, the current operating status of the unit is obtained, and combined with the predicted control parameters of the unit obtained by the parameter prediction model based on machine learning algorithm, the target adjustment strategy is generated and applied to the unit, thereby achieving intelligent control of the key control loops of the unit, flexible adjustment of parameters, and maximizing the benefits of the unit.

[0071] This invention provides an optimization method for improving the flexible operation of thermal power units. Based on the first operating status data, the operating status of the current target unit is analyzed to obtain the unit operating status result, including:

[0072] Obtain the operating data threshold table for each operating system within the target unit;

[0073] Based on the running data threshold table, the first running status data of each running system is compared with the corresponding running data threshold to obtain the first running difference;

[0074] The first operational difference is compared and analyzed with a preset difference threshold to determine the level labeling result of the first operational status data;

[0075] Based on the level labeling results of the first operating status data of each operating system, calculate the system risk coefficient of the current operating system;

[0076] The formula for calculating the system risk coefficient is as follows:

[0077] In the formula, X represents the system risk coefficient of the currently running system; This represents the number of running state data at the i-th state level within the currently running system, where... ; This represents the total number of running status data points of the currently running system; It is represented as the first operational difference between the j-th operational state data at the i-th state level in the current operating system and the corresponding operational data threshold, where, ; This represents the risk impact weight of the j-th operating state data at the i-th state level within the current operating system;

[0078] The system risk coefficients of each operating system of the target unit are set to generate the unit's operating status results.

[0079] In this embodiment, the target unit refers to the complete set of equipment and systems that use thermal power to generate electricity; the first operating status data includes power generation, efficiency, pressure and temperature, operating load, fuel consumption and emissions, unit flow rate and operating time; the operating system includes the boiler system, turbine system, generator system, auxiliary system and control system; the operating data threshold table refers to the safety threshold of the operating status data of each operating system, which is a pre-established table based on the analysis of unit design parameters, safety specifications and historical fault data.

[0080] In this embodiment, the first operating difference refers to the difference between the first operating status data and the corresponding operating data threshold; the preset difference threshold is set in advance; the level labeling result refers to the operating status level of the first operating status data, including level one, level two, level three and normal; the system risk coefficient is used to characterize the degree of abnormality in the operating status of the current target unit's operating system.

[0081] In this embodiment, the first operational difference is compared and analyzed with a preset difference threshold to determine the level labeling result of the first operational status data, including:

[0082] If the first running difference is less than the preset difference threshold and the first running status data is greater than the corresponding running data threshold, then the current first running status data is marked as second-level status data.

[0083] If the first running difference is less than the preset difference threshold and the first running status data is not greater than the corresponding running data threshold, then the current first running status data is marked as level three status data.

[0084] If the first running difference is not less than the preset difference threshold, and the first running status data is greater than the corresponding running data threshold, then the current first running status data is marked as level 1 status data.

[0085] If the first running difference is not less than the preset difference threshold and the first running status data is not greater than the corresponding running data threshold, then the current first running status data is marked as normal status data.

[0086] The beneficial effects of the above technical solution are: by analyzing the unit's operating data in real time, it helps to determine the safety operation risks and performance of each operating system of the unit, and provides data support for subsequent performance optimization and improvement, and maintenance costs.

[0087] This invention provides an optimization method for improving the flexible operation of thermal power units. It utilizes a parameter prediction model constructed using machine learning algorithms to predict key control parameters of the target unit based on first operating state data, obtaining control parameter prediction results, including:

[0088] Historical operating status data, equipment parameters, and environmental condition data of the target unit are extracted and preprocessed to obtain the first dataset.

[0089] The first dataset is divided according to a set ratio to obtain a training dataset and a test dataset;

[0090] Extract the correlation and influence features of the key control parameters of the target unit, and use the training dataset as input to train the neural network to obtain an initial prediction model;

[0091] The initial prediction model was evaluated using the test dataset, and the model was tuned based on the evaluation results to obtain the parameter prediction model.

[0092] The first operating state data is input into the parameter prediction model to obtain the control parameter prediction results.

[0093] In this embodiment, the equipment parameters include operating status parameters, such as switch status and operating mode; performance parameters, such as efficiency and speed; configuration parameters; maintenance parameters; and fault and alarm parameters. Environmental condition data refers to temperature, humidity, vibration, and dust concentration, etc. Data preprocessing includes data cleaning, noise reduction, and normalization, the purpose of which is to ensure data quality and improve model performance.

[0094] In this embodiment, the first dataset is a dataset obtained by combining the historical operating status data, equipment parameters and environmental condition data of the target unit after data preprocessing; the set ratio is set in advance.

[0095] In this embodiment, key control parameters include main steam temperature, pressure, intermediate point temperature, denitrification efficiency, etc.; correlation influence features refer to feature variables that have a significant impact on the flexible adjustment of key control parameters; the initial prediction model is a model obtained by training a neural network using the correlation influence features of the key control parameters of the target unit and the training dataset as input; the parameter prediction model is a model obtained by optimizing the initial prediction model and is used to predict the future values ​​of the key control parameters of the unit; the control parameter prediction result refers to the predicted value of the key control parameters output after the first operating state data is input into the parameter prediction model.

[0096] The beneficial effects of the above technical solution are: by using machine learning algorithms to build a parameter prediction model, it is possible to accurately predict the key control parameters of the target unit, providing data support for the flexible adjustment of subsequent key control parameters and reducing operating costs.

[0097] This invention provides an optimization method for improving the flexible operation of thermal power units. The method combines the predicted control parameters with the unit's operating status results for analysis, generating a target adjustment strategy that is applied to the target unit. This achieves flexible adjustment of the target unit and maximizes unit efficiency. The method includes:

[0098] Based on the system risk coefficient in the unit operation status results, determine the safety risk level of each operating system at the current moment;

[0099] From the risk level-adjustment strategy database, extract the first state data of each operating system at the current moment - adjustment strategy reference table based on the security risk level;

[0100] From the first state data-adjustment strategy reference table, determine the data adjustment strategy for the first operating state data labeled at different levels within the corresponding operating system, and output it as a reference adjustment strategy;

[0101] Determine the control input parameters for the key control parameters controlling the target unit;

[0102] Parameters that are directly related to the unit's economic performance are considered as associated input parameters, while parameters that are not directly related to the unit's economic performance are considered as non-associated input parameters.

[0103] Based on the reference adjustment strategy, obtain the corresponding parameter adjustment strategy for each key control parameter and the comparison parameter for the non-correlated input parameter;

[0104] Based on the predicted control parameters, the target prediction results of the associated and non-associated input parameters of the key control parameters are obtained by using a target optimization algorithm.

[0105] By combining the adjustment strategies of the associated and non-associated input parameters of each key control parameter with the target prediction results, a target adjustment strategy is generated.

[0106] In this embodiment, the system risk coefficient is used to characterize the degree of abnormality in the operating status of the current target unit's operating system; the safety risk level refers to different levels divided according to the severity of the risk, used to describe and assess potential safety threats or hazards.

[0107] In this embodiment, the risk level-adjustment strategy database consists of the safety risk level of the operating coefficient and the corresponding first state data-adjustment strategy reference table. The first state data-adjustment strategy reference table consists of the first operating state data and the corresponding data adjustment strategy. The reference adjustment strategy refers to the data adjustment strategy of the first operating state data with different level labels extracted from the first state data-adjustment strategy reference table.

[0108] In this embodiment, control input parameters refer to parameters that can affect and control the values ​​of key control parameters, such as fuel quantity, water supply, water injection quantity, etc.; associated input parameters refer to control input parameters that are directly related to the unit's economy, such as fuel quantity; and non-associated input parameters refer to control input parameters that are not directly related to the unit's economy.

[0109] In this embodiment, the reference parameter adjustment strategy refers to the data adjustment strategy for associated and unassociated input parameters extracted from the reference adjustment strategy; the target optimization algorithm refers to a pre-set method for finding the optimal solution for the future predicted input quantities of associated and unassociated input parameters of key control parameters, such as genetic optimization algorithm or particle swarm optimization algorithm; the target adjustment strategy is used to adjust associated input parameters that are directly related to the economics of key control parameters, as well as unassociated input parameters that are not directly related to economics, in order to achieve precise control of key control parameters, ensure intelligent control of the key control loop of the target unit and maximize the unit's benefits, wherein the key control loop includes main steam temperature, pressure, intermediate point temperature, denitrification, etc.

[0110] The beneficial effects of the above technical solution are: by combining the predicted results of control parameters with the results of unit operation status to generate a target adjustment strategy and apply it to the target unit, it is possible to achieve intelligent control of key control loops with the goal of optimal economy while ensuring the safe operation of the unit and equipment.

[0111] This invention provides an optimization method for improving the flexible operation of thermal power units. Based on the control parameter prediction results, a target optimization algorithm is used to analyze and obtain the target prediction results of the associated and non-associated input parameters of key control parameters, including:

[0112] Obtain the technical and economic performance indicators of key control parameters;

[0113] By taking the associated and non-associated input parameters of key control parameters as input variables and the comprehensive cost of the control loop as the output variable, and combining the corresponding technical performance indicators and economic performance indicators, a control parameter-economic model is established.

[0114] Based on the aforementioned control parameter-economic model, the objective function is to minimize the unit's hourly comprehensive cost.

[0115] The input variables are used as linear constraints, and a parameter optimization prediction model is constructed using an objective optimization algorithm.

[0116] Solving the parameter optimization prediction model yields the first prediction result sequence of the associated and non-associated input parameters of the current key control parameters within a set prediction time period.

[0117] Extract the prediction result at the current sampling time from the first prediction result sequence, and output it as the target prediction result for both associated and unassociated input parameters.

[0118] In this embodiment, key control parameters include main steam temperature, pressure, intermediate point temperature, denitrification efficiency, etc., which refer to the parameters of key control loops such as main steam temperature, pressure, intermediate point temperature, and denitrification; technical performance indicators refer to key parameters used to measure equipment performance; economic performance indicators are key parameters used to measure the economic efficiency of equipment. For example, the technical performance indicators of main steam temperature include rated temperature and temperature fluctuation range, while economic performance indicators include thermal efficiency, fuel consumption, and maintenance costs.

[0119] In this embodiment, the control parameter-economic model is established by combining the performance indicators of key control parameters with associated and non-associated input parameters as input variables and the output variable as the comprehensive cost of the control loop; the hourly comprehensive cost of the unit refers to the comprehensive cost incurred by the unit during each hour of operation, including fuel costs, maintenance costs, labor costs, etc.

[0120] In this embodiment, the target optimization algorithm refers to a pre-defined method for finding the optimal solution for the future predicted input quantities of the associated and unassociated input parameters of the key control parameters, such as genetic optimization algorithm or particle swarm optimization algorithm; the parameter optimization prediction model is used to predict the parameter values ​​of the associated and unassociated input parameters of the key control parameters within a set prediction time period; the set prediction time period is pre-defined; the first prediction result sequence refers to the predicted control input quantities of the associated and unassociated input parameters of the current key control parameters within the set prediction time period, obtained by solving the parameter optimization prediction model. The associated input parameters are control input parameters directly related to the unit's economics, such as fuel quantity; the unassociated input parameters are control input parameters not directly related to the unit's economics.

[0121] The beneficial effects of the above technical solution are: by using the target optimization algorithm, the target prediction results of the associated and non-associated input parameters of the key control parameters are obtained based on the analysis of the control parameter prediction results, providing effective data support for achieving precise control of the key control parameters.

[0122] This invention provides an optimization method to improve the flexible operation of thermal power units. It combines the adjustment strategies of associated and non-associated input parameters for each key control parameter with target prediction results to generate a target adjustment strategy, including:

[0123] Based on the combined analysis of the parameter adjustment strategy and the level labeling results of the associated input parameters, the adjustment range of the first parameter is obtained;

[0124] Based on the combined analysis of the parameter adjustment strategy for non-associated input parameters and the results of level labeling, the adjustment range of the second parameter is obtained;

[0125] When the correlation-target prediction results of all associated input parameters of the key control parameter are within the adjustment range of the corresponding first parameter, the correlation-target prediction results are output as the optimized values ​​of the associated input parameters of the corresponding associated input parameters.

[0126] When the target prediction result of one or more associated input parameters of a key control parameter does not fall within the adjustment range of the first parameter, the corresponding associated input parameter is marked as the first parameter.

[0127] The linear constraint condition of the first parameter is adjusted using the constraint adjustment coefficient, and the parameter optimization prediction model is re-solved based on the adjusted linear constraint condition to obtain the target prediction results for associated and non-associated input parameters.

[0128] Analyze the target prediction results of the associated input parameters with the corresponding adjustment range of the first parameter, determine the optimized value of the associated input parameters, and output it.

[0129] If the non-associative target prediction result of the non-associative input parameter falls within the adjustment range of the second parameter, then the current non-associative target prediction result is used as the optimized value of the non-associative input parameter.

[0130] If the non-associative target prediction result of the current non-associative input parameter does not fall within the adjustment range of the second parameter, then calculate the absolute difference between the non-associative target prediction result and the upper limit and lower limit of the second parameter adjustment respectively, and obtain the first absolute difference and the second absolute difference accordingly;

[0131] If the first absolute difference is greater than the second absolute difference, then the upper limit of the second parameter adjustment is used as the optimization value of the non-associative input parameter;

[0132] Otherwise, adjust the lower limit of the second parameter as the optimized value for the non-associative input parameter;

[0133] Based on the optimized values ​​of the associated input parameters and the optimized values ​​of the non-associated input parameters for each key control parameter, a target adjustment strategy is generated.

[0134] In this embodiment, the level labeling results include Level 1, Level 2, Level 3, and Normal Level; the first parameter adjustment range refers to the allowed range of parameter values ​​for associated input parameters; the second parameter adjustment range refers to the allowed range of parameter values ​​for non-associated input parameters; the association-prediction result refers to the predicted amount of the associated input parameters at the current sampling time, obtained by analyzing the prediction results of control parameters using a target optimization algorithm.

[0135] In this embodiment, the optimized value of the associated input parameter refers to the best value of the associated input parameter that satisfies the prediction results of the key control parameters, with the safety of the unit and equipment as the boundary and the best economic efficiency as the objective; the first parameter refers to the associated input parameter whose correlation-target prediction result does not fall within the adjustment range of the first parameter; the constraint adjustment coefficient is used to adjust and optimize the linear constraint conditions; the non-associated-prediction result refers to the predicted amount of the non-associated input parameter at the current sampling time, obtained by analyzing the non-associated input parameter using the target optimization algorithm based on the control parameter prediction result; the optimized value of the non-associated input parameter refers to the best value of the non-associated input parameter that satisfies the prediction results of the key control parameters, with the safety of the unit and equipment as the boundary and the best economic efficiency as the objective.

[0136] In this embodiment, the first absolute difference refers to the absolute difference between the non-correlated prediction result and the second parameter adjustment upper limit, wherein the second parameter adjustment upper limit refers to the maximum allowed value of the parameter; the second absolute difference refers to the absolute difference between the non-correlated prediction result and the second parameter adjustment lower limit, wherein the second parameter adjustment lower limit refers to the minimum allowed value of the parameter.

[0137] The beneficial effects of the above technical solution are: by combining the adjustment strategy of the associated input parameters and the comparison parameters of the non-associated input parameters of the key control parameters with the target prediction results determined by the target optimization algorithm, the target adjustment strategy is generated and applied to the unit. This can effectively ensure that, on the basis of safe operation of the unit and equipment, the unit achieves intelligent and flexible control of the key control loops of the unit with the goal of optimal economy.

[0138] This invention provides an optimization method for improving the flexible operation of thermal power units. The formula for calculating the constraint adjustment coefficient is shown below:

[0139] In the formula, P represents the constraint optimization coefficient; Represented as a linear constraint adjustment factor; This represents the upper limit of the adjustment for the corresponding first parameter; This represents the lower limit of the adjustment of the corresponding first parameter; It is represented as the absolute difference between the current first parameter's correlation-target prediction result and the corresponding first parameter's adjustment upper limit; It is represented as the absolute difference between the current first parameter's correlation-target prediction result and the corresponding first parameter's lower limit; This represents the weight of the current first parameter's impact on unit safety. This represents the contribution compensation coefficient of the current first parameter to the unit's economic efficiency.

[0140] The beneficial effects of the above technical solution are: by calculating the constraint adjustment coefficient, the linear constraint conditions of the associated input parameters whose correlation-target prediction results do not fall within the adjustment range of the corresponding first parameter are adjusted, which helps to improve the performance of the parameter optimization prediction model and achieve precise control of key control parameters.

[0141] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An optimization method for improving flexible operation of a thermal power unit, characterized in that, The method comprises the following steps: Step 1: Real-time monitoring of the running state of the target unit to obtain first running state data; Step 2: Analyzing the running state of the current target unit based on the first running state data to obtain a unit running state result; Step 3: Using a parameter prediction model constructed by a machine learning algorithm to predict the key control parameters of the target unit based on the first running state data to obtain a control parameter prediction result; Step 4: Combining and analyzing the control parameter prediction result and the unit running state result to generate a target adjustment strategy to act on the target unit to maximize the unit benefit.

2. The optimization method for improving flexible operation of a thermal power unit according to claim 1, characterized in that, The first running state data includes power generation, efficiency, pressure and temperature, operating load, fuel consumption and emissions, unit flow, and operating time.

3. The optimization method for improving flexible operation of a thermal power unit according to claim 1, characterized in that, Based on the first running state data, the running state of the current target unit is analyzed to obtain a unit running state result, which includes: Obtaining the running data threshold table of each running system in the target unit; Comparing the first running state data of each running system with the corresponding running data threshold based on the running data threshold table to obtain a first running difference; Comparing and analyzing the first running difference with a preset difference threshold to determine the grade labeling result of the first running state data; Based on the grade labeling result of the first running state data of each running system, the system risk coefficient of the current running system is calculated. The formula for calculating the system risk coefficient is as follows: ; wherein X represents a system risk coefficient of a current running system; the number of running state data representing the i-th state level within the current running system, wherein, ； the total number of the running state data representing the running state of the current running system; a first operating difference value representing a difference between the jth operating state data of the i state level within the current operating system and the corresponding operating data threshold value, wherein ； represents the risk influence weight of the jth running state data of the ith state grade in the current running system; The system risk coefficients of each running system of the target unit are collected to generate a unit running state result.

4. The optimization method for improving flexible operation of a thermal power unit according to claim 1, characterized in that, Using a parameter prediction model constructed by a machine learning algorithm, the key control parameters of the target unit are predicted based on the first running state data to obtain a control parameter prediction result, which includes: Extracting the historical running state data, equipment parameters, and environmental condition data of the target unit, and after data preprocessing, a first data set is obtained; Dividing the first data set according to a set proportion to obtain a training data set and a test data set; Extracting the associated influence features of the key control parameters of the target unit, and combining the training data set as an input quantity to train a neural network to obtain an initial prediction model; Using the test data set to evaluate the performance of the initial prediction model, and based on the performance evaluation result, the model is optimized to obtain a parameter prediction model; Inputting the first running state data into the parameter prediction model to obtain a control parameter prediction result.

5. The optimization method for improving flexible operation of a thermal power unit according to claim 1, characterized in that, Combining and analyzing the control parameter prediction result and the unit running state result to generate a target adjustment strategy to act on the target unit to maximize the unit benefit, which includes: According to the system risk coefficient in the unit running state result, the safety risk level of each running system at the current time is determined; From the risk level-adjustment strategy database, based on the safety risk level, the first state data-adjustment strategy reference table of each running system at the current time is extracted; From the first state data-adjustment strategy reference table, determine the data adjustment strategy corresponding to the first running state data marked with different levels in the running system, and output as the reference adjustment strategy; Determine the control input parameters of the key control parameters of the target unit; Regarding the parameters directly related to the economy of the unit as associated input parameters, and the parameters not directly related to the economy of the unit as non-associated input parameters; According to the reference adjustment strategy, obtain the control parameter adjustment strategy of the associated input parameters and the non-associated input parameters of each key control parameter respectively; Based on the control parameter prediction result, the target prediction result of the associated input parameters and the non-associated input parameters of the key control parameters is obtained by using a target optimization algorithm; The target adjustment strategy is generated by combining and analyzing the control parameter adjustment strategy and the target prediction result of the associated input parameters and the non-associated input parameters of each key control parameter.

6. The optimization method for improving flexible operation of a thermal power unit according to claim 5, characterized in that, Based on the control parameter prediction result, the target prediction result of the associated input parameters and the non-associated input parameters of the key control parameters is obtained by using a target optimization algorithm, including: Obtain the technical performance index and economic performance index of the key control parameters; Take the associated input parameters and the non-associated input parameters of the key control parameters as input variables, and take the control loop comprehensive cost as output variable, and combine the corresponding technical performance index and economic performance index to establish a control parameter-economy model; Based on the control parameter-economy model, taking the minimum unit hour-level comprehensive cost as the objective function, Taking the input variable as a linear constraint condition, a parameter optimization prediction model is constructed by using a target optimization algorithm; Solve the parameter optimization prediction model to obtain the first prediction result sequence of the associated input parameters and the non-associated input parameters of the current key control parameters within a set prediction time period; Extract the prediction result at the current sampling time from the first prediction result sequence as the target prediction result of the associated input parameters and the non-associated input parameters.

7. The optimization method for improving flexible operation of a thermal power unit according to claim 5, characterized in that, The target adjustment strategy is generated by combining and analyzing the control parameter adjustment strategy and the target prediction result of the associated input parameters and the non-associated input parameters of each key control parameter, including: According to the combination analysis of the control parameter adjustment strategy of the associated input parameters and the level marking result, a first parameter adjustment range is obtained; According to the combination analysis of the control parameter adjustment strategy of the non-associated input parameters and the level marking result, a second parameter adjustment range is obtained; When the associated-target prediction results of all associated input parameters of the key control parameters belong to the corresponding first parameter adjustment range, the associated-target prediction result is output as the associated input parameter optimization value of the corresponding associated input parameter; When the associated-target prediction results of one or more associated input parameters of the key control parameters do not belong to the first parameter adjustment range, the corresponding associated input parameter is marked as a first parameter; Adjust the linear constraint condition of the first parameter by using a constraint adjustment coefficient, and based on the adjusted linear constraint condition, re-solve the parameter optimization prediction model to obtain the target prediction result of the associated input parameters and the non-associated input parameters; The target prediction result of the associated input parameter is analyzed with the corresponding first parameter adjustment range to determine the associated input parameter optimization value and output the associated input parameter optimization value; If the non-associated-target prediction result of the non-associated input parameter belongs to the second parameter adjustment range, the current non-associated-target prediction result is taken as the non-associated input parameter optimization value; If the non-associated-target prediction result of the current non-associated input parameter does not belong to the second parameter adjustment range, the parameter absolute difference between the non-associated-target prediction result and the upper limit of the second parameter adjustment range and the lower limit of the second parameter adjustment range is calculated respectively, and the first absolute difference and the second absolute difference are obtained correspondingly; If the first absolute difference is greater than the second absolute difference, the upper limit of the second parameter adjustment range is taken as the non-associated input parameter optimization value; Otherwise, the lower limit of the second parameter adjustment range is taken as the non-associated input parameter optimization value; Based on the associated input parameter optimization value of the associated input parameter of each key control parameter and the non-associated input parameter optimization value of the non-associated input parameter, a target adjustment strategy is generated.

8. The optimization method for improving flexible operation of a thermal power unit according to claim 7, characterized in that, The calculation formula of the constraint adjustment coefficient is as follows: ; wherein P represents a constraint optimization coefficient; denoted as linear constraint adjustment factors; representing a corresponding first parameter adjustment upper limit for the current first parameter; a corresponding first parameter adjustment lower limit expressed as a current first parameter; an absolute difference between the association-target prediction result expressed as a current first parameter and an upper limit of the corresponding first parameter adjustment; an absolute difference between the associated-target prediction result expressed as a current first parameter and a lower limit of the first parameter adjustment; representing the influence weight of the current first parameter on the safety of the unit; The current first parameter is represented as a compensation coefficient of the contribution to the economic efficiency of the unit.

Images