Multi-working-condition fire coal combustion control system and method based on machine learning

Abstract

The invention relates to the technical field of fire coal combustion control, and discloses a multi-working-condition fire coal combustion control system and method based on machine learning, and the system comprises an obtaining module which is used for determining a plurality of working condition types, obtaining a plurality of historical combustion logs, carrying out the feature extraction of the historical combustion data of the historical combustion logs, and obtaining a feature extraction result; obtaining combustion state data of each working condition category; the construction module is used for constructing a combustion state evaluation model and a combustion control model, and generating a real-time comprehensive combustion state evaluation value according to the combustion state evaluation model; the control module is used for judging whether optimization control is needed or not according to the real-time comprehensive combustion state evaluation value, and if yes, a first combustion control strategy is generated based on a combustion control model and a control instruction is issued; and the correction module is used for obtaining the real-time combustion state data based on the preset feedback time node and judging whether correction is carried out or not, if yes, a correction instruction is generated based on a preset correction reference library, and the fire coal combustion efficiency and the control efficiency are guaranteed.

Description

Technical Field

[0001] The present application relates to the field of coal combustion control technology, and in particular to a multi-condition coal combustion control system and method based on machine learning. Background Art

[0002] The coal combustion control process is a complex multivariable system that requires monitoring based on multiple combustion parameters and applying corresponding control measures based on the monitoring results. The effectiveness of the control depends largely on the accuracy of the monitoring results, the timeliness of the control, and the rationality of the control.

[0003] In existing technologies, training mostly relies on massive combustion data as samples to achieve predictions of combustion efficiency and pollutant emissions. However, the data volume is large, the prediction results are not accurate enough, and accurate control parameters cannot be provided in a timely manner, which cannot guarantee the control effect and reduce the coal combustion efficiency and control efficiency. Summary of the Invention

[0004] In order to solve the above technical problems, the present application provides a multi-condition coal-fired combustion control system and method based on machine learning. By determining the combustion state data of each working condition category and constructing a combustion state evaluation model and a combustion control model, the combustion state is accurately evaluated and the first coal-fired control strategy is formulated in a timely manner. The application effect is evaluated according to the coal-fired state data after the first coal-fired control strategy, and the shortcomings of the first coal-fired control strategy are discovered and adjusted in a timely manner to ensure coal combustion efficiency and control efficiency.

[0005] In some embodiments of the present application, a multi-condition coal combustion control system and method based on machine learning is provided, including: an acquisition module, configured to determine a plurality of operating condition categories and obtain a plurality of historical combustion logs for each operating condition category, perform feature extraction on the historical combustion data in the plurality of historical combustion logs, and obtain combustion state data for each operating condition category; A construction module is used to construct a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generate a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; a control module, configured to determine whether optimization control is required based on the real-time comprehensive combustion state evaluation value, and if so, generate a first combustion control strategy based on the combustion control model and issue a control instruction; The correction module is used to obtain real-time combustion status data based on a preset feedback time node and determine whether to perform correction. If so, it generates correction instructions based on a preset correction reference library.

[0006] In some embodiments of the present application, feature extraction is performed on historical combustion data in a number of historical combustion logs to obtain combustion state data for each operating condition category, including: Obtain historical combustion logs for all operating conditions and perform cluster analysis to obtain multiple operating condition categories and several historical combustion logs for each operating condition category; Preset a number of combustion evaluation indicators, and set each combustion evaluation indicator to strongly correlate historical combustion data in a number of historical combustion logs for each operating condition category; A time reference line is established based on the historical monitoring duration of each historical combustion log of each operating condition category, and a collection node is established based on a preset time interval. Multiple sampling intensity-related historical combustion data and other historical combustion data in the corresponding historical combustion log are collected according to the collection node; Generate an initial combustion state assessment value of the corresponding combustion evaluation index based on all strongly correlated historical combustion data at the same acquisition node; Mapping the initial combustion state evaluation values of all combustion evaluation indicators at multiple acquisition nodes and multiple historical combustion data onto the time reference line of the corresponding historical combustion log to obtain multiple initial combustion state evaluation value change curves and multiple historical combustion data change curves; Perform dependency analysis on each historical combustion data change curve in the same historical combustion log and a number of initial combustion state assessment value change curves to obtain the dependency degree; A correlation coefficient is set between each historical combustion data change curve and the corresponding initial combustion state evaluation value change curve according to the degree of dependence. If the correlation coefficient is greater than a preset correlation coefficient threshold, the corresponding historical combustion data is set as the associated historical combustion data of the corresponding combustion evaluation index. Comparing and analyzing the associated historical combustion data of the same combustion evaluation index determined in each historical combustion log of the same operating condition category to obtain the occurrence ratio of the same associated historical combustion data, and eliminating the associated historical combustion data with an occurrence ratio less than a preset ratio threshold; The strongly correlated historical combustion data of each combustion evaluation index of the same operating condition category and the remaining correlated historical combustion data are set as the combustion state data of the corresponding operating condition category, and a weight coefficient of each combustion state data is set.

[0007] In some embodiments of the present application, a combustion state assessment model is constructed based on the combustion state data of each operating condition category, including: Obtaining combustion state data for each operating condition category at a preset historical time node, and generating combustion state evaluation values for a number of combustion evaluation indicators at the corresponding preset historical time nodes; Generate a comprehensive combustion state evaluation value at a corresponding historical time node based on the combustion state evaluation values of several combustion evaluation indicators at the same preset historical time node; The combustion state data at multiple preset historical time nodes of each operating condition category are used as training input data, and the comprehensive combustion state evaluation values at the corresponding preset historical time nodes are used as training output data. Neural network training is performed to obtain a combustion state evaluation model.

[0008] In some embodiments of the present application, the combustion control model includes: Divide multiple preset historical time nodes into several analysis periods according to preset time intervals; Analyze the comprehensive combustion state evaluation value of each operating condition category in each analysis period, and determine the change value and change trend of the comprehensive combustion state evaluation value in the corresponding analysis period; If the change trend is normal and the change value is greater than the preset change value threshold, the period in the analysis period that is in the normal trend is intercepted and set as the focus period; If the final comprehensive combustion state evaluation value of the focus period is greater than the preset evaluation value threshold, a number of historical control parameters in the corresponding focus period are retrieved; Obtaining historical control adjustment values of the retrieved historical control parameters in the corresponding focus period, and constructing a historical control parameter-adjustment value mapping table for the corresponding focus period; Mapping the historical control parameter-adjustment value mapping table again with the combustion state data at the first preset historical time node of the corresponding focus period to obtain a combustion state data-historical control parameter-adjustment value mapping table; The combustion state data in a plurality of combustion state data-historical control parameter-control value mapping tables are used as training input data, and the historical control parameter-control value mapping tables corresponding to the combustion state data are used as training output data to perform neural network training and obtain a combustion control model.

[0009] In some embodiments of the present application, determining whether optimization control is required based on the real-time comprehensive combustion state evaluation value includes: Determine the real-time operating condition category and obtain the real-time combustion state data at the current monitoring time node, input the real-time operating condition category and the real-time combustion state data into the combustion state evaluation model to obtain a real-time comprehensive combustion state evaluation value; Constructing a real-time combustion state data change curve according to the real-time combustion state data at multiple monitoring time nodes in the current monitoring cycle; Preset several monitoring time nodes in the future period, perform curve trend extrapolation on the real-time combustion state data change curve, obtain predicted combustion state data at several monitoring time nodes in the future period, and input the data into the combustion state assessment model to obtain predicted comprehensive combustion state assessment values at several monitoring time nodes in the future period; generating a compensation coefficient for the real-time comprehensive combustion state evaluation value at the current monitoring time node according to the change characteristics of the predicted comprehensive combustion state evaluation values at several monitoring time nodes in the future period; The real-time comprehensive combustion state evaluation value is corrected according to the compensation coefficient. If the corrected real-time comprehensive combustion state evaluation value is less than the preset evaluation value, it is judged that optimization control is required. If the corrected real-time comprehensive combustion state evaluation value is greater than the preset evaluation value, it is judged that optimization control is not required.

[0010] In some embodiments of the present application, generating a first combustion control strategy based on a combustion control model and issuing a control instruction includes: If optimization control is required, the real-time combustion state data at the current monitoring time node is input into the combustion control model to obtain the control values of several control parameters and generate a first combustion control strategy; Generate a standard change rate and a standard change value for each combustion state data after the first combustion control strategy, generate a standard change model for the corresponding combustion state data according to the standard change rate and the standard change value, and set a number of preset feedback time nodes; A control instruction is generated according to the first combustion control strategy, and a monitoring instruction is generated according to a preset feedback time node.

[0011] In some embodiments of the present application, obtaining real-time combustion state data based on a preset feedback time node and determining whether to perform corrections include: Acquire real-time combustion state data based on preset feedback time nodes, and calculate the actual change rate and actual change value at the corresponding preset feedback time nodes; The actual change rate and actual change value of each real-time combustion state data at the preset feedback time node are input into the standard change model to obtain the difference degree; Generate a comprehensive difference degree of the corresponding preset feedback time node according to the difference degree of each real-time combustion state data at the same preset feedback time node, and construct a comprehensive difference degree sequence according to the time sequence of the preset feedback time nodes; If the comprehensive difference degree sequence does not show a downward trend, it is determined that the first combustion control strategy needs to be modified.

[0012] In some embodiments of the present application, the preset correction reference library includes: A plurality of preset differences are set according to a standard variation model of each real-time combustion state data, and each preset difference is mapped to a plurality of preset control parameters to be corrected and corresponding preset correction values; Randomly combining multiple preset difference degrees of different real-time combustion state data to obtain multiple preset comprehensive difference degrees; Perform conflict analysis based on a number of preset control parameters to be modified and their corresponding preset correction values mapped to the same preset comprehensive difference, and determine whether to perform adjustment and optimization based on the analysis results; If so, generating a preset correction strategy corresponding to the preset comprehensive difference according to the optimized plurality of preset control parameters to be corrected and the corresponding preset correction values; A preset correction reference library is constructed based on multiple preset comprehensive differences and corresponding preset correction strategies.

[0013] In some embodiments of the present application, generating a correction instruction based on a preset correction reference library includes: Obtaining the comprehensive difference at the current preset feedback time node, and performing similarity analysis with the preset comprehensive difference to obtain multiple similarities; The preset correction strategy mapped to the preset comprehensive difference degree with the greatest similarity is set as the correction strategy of the first combustion control strategy, and a correction instruction is generated.

[0014] In some embodiments of the present application, a multi-condition coal combustion control method based on machine learning is also included: Determine multiple operating condition categories and obtain several historical combustion logs for each operating condition category, perform feature extraction on the historical combustion data in the several historical combustion logs, and obtain combustion state data for each operating condition category; Construct a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generate a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; Determine whether optimization control is required based on the real-time comprehensive combustion state evaluation value, and if so, generate a first combustion control strategy based on the combustion control model and issue a control instruction; The real-time combustion status data is obtained based on the preset feedback time node, and it is determined whether correction is required. If so, a correction instruction is generated based on the preset correction reference library.

[0015] Compared with the prior art, the multi-condition coal combustion control system and method based on machine learning in the embodiments of the present application have the following advantages: By determining the combustion state data of each operating condition category and constructing a combustion state evaluation model and a combustion control model, the combustion state is accurately evaluated and the first coal control strategy is formulated in a timely manner. The application effect is evaluated based on the coal state data after the first coal control strategy is implemented, and the shortcomings of the first coal control strategy are discovered and adjusted in a timely manner to ensure coal combustion efficiency and control efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of a multi-condition coal combustion control system based on machine learning in an embodiment of the present application; Figure 2It is a flow chart of a multi-condition coal combustion control method based on machine learning in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0018] In the description of this application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application.

[0019] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout this application, unless otherwise specified, "plurality" means two or more.

[0020] In the description of this application, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.

[0021] like Figure 1 As shown, the multi-condition coal combustion control system based on machine learning in the embodiment of the present application includes: an acquisition module, configured to determine a plurality of operating condition categories and obtain a plurality of historical combustion logs for each operating condition category, perform feature extraction on the historical combustion data in the plurality of historical combustion logs, and obtain combustion state data for each operating condition category; A construction module is used to construct a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generate a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; a control module, configured to determine whether optimization control is required based on the real-time comprehensive combustion state evaluation value, and if so, generate a first combustion control strategy based on the combustion control model and issue a control instruction; The correction module is used to obtain real-time combustion status data based on a preset feedback time node and determine whether to perform correction. If so, it generates correction instructions based on a preset correction reference library.

[0022] In some embodiments of the present application, feature extraction is performed on historical combustion data in a number of historical combustion logs to obtain combustion state data for each operating condition category, including: Obtain historical combustion logs for all operating conditions and perform cluster analysis to obtain multiple operating condition categories and several historical combustion logs for each operating condition category; Preset a number of combustion evaluation indicators, and set each combustion evaluation indicator to strongly correlate historical combustion data in a number of historical combustion logs for each operating condition category; A time reference line is established based on the historical monitoring duration of each historical combustion log of each operating condition category, and a collection node is established based on a preset time interval. Multiple sampling intensity-related historical combustion data and other historical combustion data in the corresponding historical combustion log are collected according to the collection node; Generate an initial combustion state assessment value of the corresponding combustion evaluation index based on all strongly correlated historical combustion data at the same acquisition node; Mapping the initial combustion state evaluation values of all combustion evaluation indicators at multiple acquisition nodes and multiple historical combustion data onto the time reference line of the corresponding historical combustion log to obtain multiple initial combustion state evaluation value change curves and multiple historical combustion data change curves; Perform dependency analysis on each historical combustion data change curve in the same historical combustion log and a number of initial combustion state assessment value change curves to obtain the dependency degree; A correlation coefficient is set between each historical combustion data change curve and the corresponding initial combustion state evaluation value change curve according to the degree of dependence. If the correlation coefficient is greater than a preset correlation coefficient threshold, the corresponding historical combustion data is set as the associated historical combustion data of the corresponding combustion evaluation index. Comparing and analyzing the associated historical combustion data of the same combustion evaluation index determined in each historical combustion log of the same operating condition category to obtain the occurrence ratio of the same associated historical combustion data, and eliminating the associated historical combustion data with an occurrence ratio less than a preset ratio threshold; The strongly correlated historical combustion data of each combustion evaluation index of the same operating condition category and the remaining correlated historical combustion data are set as the combustion state data of the corresponding operating condition category, and a weight coefficient of each combustion state data is set.

[0023] In this embodiment, the combustion evaluation indicators include combustion efficiency, exhaust heat loss, power generation coal consumption, gas emissions, solid emissions, equipment safety, combustion stability and other indicators.

[0024] In this embodiment, the strongly correlated historical combustion data is set in advance and refers to data that can directly evaluate the combustion state of the corresponding combustion evaluation index, and the correlated historical combustion data refers to data that can indirectly evaluate the combustion state of the corresponding combustion evaluation index.

[0025] In this embodiment, the degree of dependence refers to the degree to which the initial combustion state evaluation value change curve changes with the change of the historical combustion data change curve. When the degree of dependence is greater, the corresponding correlation coefficient is greater, and vice versa.

[0026] In this embodiment, the occurrence ratio refers to the ratio of the number of occurrences of the same associated historical combustion data determined in different historical combustion logs of the same operating condition category to the total number of occurrences of all historical combustion logs.

[0027] In this embodiment, the weight coefficient of each combustion state data is set according to the correlation coefficient between the corresponding combustion state data and the corresponding combustion evaluation index, wherein the correlation coefficient of the strongly correlated historical combustion data is set in advance.

[0028] In this embodiment, by determining the combustion state data of each operating condition category, the foundation is laid for the subsequent construction of a combustion state evaluation model and a combustion control model, thereby improving the accuracy of combustion state evaluation for different operating condition categories, thereby timely formulating reasonable combustion control strategies and improving coal combustion efficiency.

[0029] In some embodiments of the present application, a combustion state assessment model is constructed based on the combustion state data of each operating condition category, including: Obtaining combustion state data for each operating condition category at a preset historical time node, and generating combustion state evaluation values for a number of combustion evaluation indicators at the corresponding preset historical time nodes; Generate a comprehensive combustion state evaluation value at a corresponding historical time node based on the combustion state evaluation values of several combustion evaluation indicators at the same preset historical time node; The combustion state data at multiple preset historical time nodes of each operating condition category are used as training input data, and the comprehensive combustion state evaluation values at the corresponding preset historical time nodes are used as training output data. Neural network training is performed to obtain a combustion state evaluation model.

[0030] In this embodiment, the preset historical time node is set in advance and is set according to the historical time node of the historical combustion log corresponding to the working condition category, and the preset historical time node is a continuous and representative time node.

[0031] In this embodiment, by constructing combustion state evaluation models for different operating conditions, a foundation is laid for subsequent judgment on whether to perform optimization control, thereby improving the timeliness and rationality of multi-condition coal combustion control and ensuring coal combustion efficiency.

[0032] In some embodiments of the present application, the combustion control model includes: Divide multiple preset historical time nodes into several analysis periods according to preset time intervals; Analyze the comprehensive combustion state evaluation value of each operating condition category in each analysis period, and determine the change value and change trend of the comprehensive combustion state evaluation value in the corresponding analysis period; If the change trend is normal and the change value is greater than the preset change value threshold, the period in the analysis period that is in the normal trend is intercepted and set as the focus period; If the final comprehensive combustion state evaluation value of the focus period is greater than the preset evaluation value threshold, a number of historical control parameters in the corresponding focus period are retrieved; Obtaining historical control adjustment values of the retrieved historical control parameters in the corresponding focus period, and constructing a historical control parameter-adjustment value mapping table for the corresponding focus period; Mapping the historical control parameter-adjustment value mapping table again with the combustion state data at the first preset historical time node of the corresponding focus period to obtain a combustion state data-historical control parameter-adjustment value mapping table; The combustion state data in a plurality of combustion state data-historical control parameter-control value mapping tables are used as training input data, and the historical control parameter-control value mapping tables corresponding to the combustion state data are used as training output data to perform neural network training and obtain a combustion control model.

[0033] In this embodiment, the preset time interval refers to the maximum control time corresponding to the satisfaction of the combustion control efficiency, that is, if the change trend of the comprehensive combustion state evaluation value is a normal trend, the change value is greater than the preset change value threshold, and the final comprehensive combustion state evaluation value of the focus period is greater than the preset evaluation value threshold, that is, within the control time required to meet the combustion control efficiency, the comprehensive combustion state evaluation value is greater than the preset evaluation value threshold, which greatly improves the combustion efficiency.

[0034] In this embodiment, by retrieving the historical control parameters in the corresponding time period of interest and calculating the control values of the corresponding historical control parameters, the historical control parameters that meet the coal combustion efficiency are screened out, and the combustion control model is constructed in combination with the combustion state data, laying the foundation for the subsequent formulation of the first combustion control strategy and improving the combustion control efficiency and accuracy.

[0035] In some embodiments of the present application, determining whether optimization control is required based on the real-time comprehensive combustion state evaluation value includes: Determine the real-time operating condition category and obtain the real-time combustion state data at the current monitoring time node, input the real-time operating condition category and the real-time combustion state data into the combustion state evaluation model to obtain a real-time comprehensive combustion state evaluation value; Constructing a real-time combustion state data change curve according to the real-time combustion state data at multiple monitoring time nodes in the current monitoring cycle; Preset several monitoring time nodes in the future period, perform curve trend extrapolation on the real-time combustion state data change curve, obtain predicted combustion state data at several monitoring time nodes in the future period, and input the data into the combustion state assessment model to obtain predicted comprehensive combustion state assessment values at several monitoring time nodes in the future period; generating a compensation coefficient for the real-time comprehensive combustion state evaluation value at the current monitoring time node according to the change characteristics of the predicted comprehensive combustion state evaluation values at several monitoring time nodes in the future period; The real-time comprehensive combustion state evaluation value is corrected according to the compensation coefficient. If the corrected real-time comprehensive combustion state evaluation value is less than the preset evaluation value, it is judged that optimization control is required. If the corrected real-time comprehensive combustion state evaluation value is greater than the preset evaluation value, it is judged that optimization control is not required.

[0036] In this embodiment, the change characteristics refer to the change trend, fluctuation degree and change rate of the predicted comprehensive combustion state evaluation value. If the change trend is a downward trend, the greater the fluctuation degree and the faster the change rate, the smaller the compensation coefficient, and vice versa. The value range of the compensation coefficient is (0.75, 1.25).

[0037] In this embodiment, the accuracy of comprehensive judgment on coal combustion efficiency at the current monitoring time node is improved by using the real-time comprehensive combustion state evaluation value and the compensation coefficient of the predicted comprehensive combustion state evaluation value, thereby ensuring control timeliness and control efficiency.

[0038] In some embodiments of the present application, generating a first combustion control strategy based on a combustion control model and issuing a control instruction includes: If optimization control is required, the real-time combustion state data at the current monitoring time node is input into the combustion control model to obtain the control values of several control parameters and generate a first combustion control strategy; Generate a standard change rate and a standard change value for each combustion state data after the first combustion control strategy, generate a standard change model for the corresponding combustion state data according to the standard change rate and the standard change value, and set a number of preset feedback time nodes; A control instruction is generated according to the first combustion control strategy, and a monitoring instruction is generated according to a preset feedback time node.

[0039] In this embodiment, the standard change rate and the standard change value are determined based on the historical control logs of the control values of several control parameters in the first combustion control strategy.

[0040] In some embodiments of the present application, obtaining real-time combustion state data based on a preset feedback time node and determining whether to perform corrections include: Acquire real-time combustion state data based on preset feedback time nodes, and calculate the actual change rate and actual change value at the corresponding preset feedback time nodes; The actual change rate and actual change value of each real-time combustion state data at the preset feedback time node are input into the standard change model to obtain the difference degree; Generate a comprehensive difference degree of the corresponding preset feedback time node according to the difference degree of each real-time combustion state data at the same preset feedback time node, and construct a comprehensive difference degree sequence according to the time sequence of the preset feedback time nodes; If the comprehensive difference degree sequence does not show a downward trend, it is determined that the first combustion control strategy needs to be modified.

[0041] In this embodiment, the degree of difference is set based on the difference between the actual change rate and actual change value at the same preset feedback time node and the standard change rate and standard change value of the corresponding combustion state data. The larger the difference, the greater the degree of difference, and vice versa.

[0042] In this embodiment, the comprehensive difference is calculated based on the difference of each real-time combustion state data and the corresponding weight coefficient.

[0043] In this embodiment, the downward trend includes a long-term decreasing trend in the comprehensive difference. If the comprehensive difference sequence does not show a downward trend, but shows a cyclic fluctuation, an upward trend, a horizontal trend, etc., it means that there is a large difference between the change value or change rate of the real-time combustion state data and the standard change value or standard change rate, which reduces the coal combustion efficiency and control effect.

[0044] In some embodiments of the present application, the preset correction reference library includes: A plurality of preset differences are set according to a standard variation model of each real-time combustion state data, and each preset difference is mapped to a plurality of preset control parameters to be corrected and corresponding preset correction values; Randomly combining multiple preset difference degrees of different real-time combustion state data to obtain multiple preset comprehensive difference degrees; Perform conflict analysis based on a number of preset control parameters to be modified and their corresponding preset correction values mapped to the same preset comprehensive difference, and determine whether to perform adjustment and optimization based on the analysis results; If so, generating a preset correction strategy corresponding to the preset comprehensive difference according to the optimized plurality of preset control parameters to be corrected and the corresponding preset correction values; A preset correction reference library is constructed based on multiple preset comprehensive differences and corresponding preset correction strategies.

[0045] In this embodiment, the preset comprehensive difference is obtained by randomly selecting a plurality of preset differences of each real-time combustion state data, obtaining a preset difference of all real-time combustion state data, and calculating the difference.

[0046] In this embodiment, the preset control parameter to be corrected and the corresponding preset correction value are obtained based on historical correction logs.

[0047] In some embodiments of the present application, generating a correction instruction based on a preset correction reference library includes: Obtaining the comprehensive difference at the current preset feedback time node, and performing similarity analysis with the preset comprehensive difference to obtain multiple similarities; The preset correction strategy mapped to the preset comprehensive difference degree with the greatest similarity is set as the correction strategy of the first combustion control strategy, and a correction instruction is generated.

[0048] In some embodiments of the present application, Figure 2 As shown, it also includes a multi-condition coal combustion control method based on machine learning: Step S201: determining multiple operating condition categories and obtaining a number of historical combustion logs for each operating condition category, performing feature extraction on the historical combustion data in the number of historical combustion logs to obtain combustion state data for each operating condition category; Step S202: constructing a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generating a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; Step S203: determining whether optimization control is required based on the real-time comprehensive combustion state evaluation value; if so, generating a first combustion control strategy based on the combustion control model and issuing a control instruction; Step S204: acquiring real-time combustion state data based on a preset feedback time node, and determining whether to perform correction; if so, generating correction instructions based on a preset correction reference library.

[0049] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and replacements can be made without departing from the technical principles of the present application. These improvements and replacements should also be regarded as the scope of protection of the present application.

Claims

1. A multi-condition coal combustion control system based on machine learning, characterized by: include: an acquisition module, configured to determine a plurality of operating condition categories and obtain a plurality of historical combustion logs for each operating condition category, perform feature extraction on the historical combustion data in the plurality of historical combustion logs, and obtain combustion state data for each operating condition category; A construction module is used to construct a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generate a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; a control module, configured to determine whether optimization control is required based on the real-time comprehensive combustion state evaluation value, and if so, generate a first combustion control strategy based on the combustion control model and issue a control instruction; The correction module is used to obtain real-time combustion status data based on a preset feedback time node and determine whether to perform correction. If so, it generates correction instructions based on a preset correction reference library.

2. The multi-condition coal combustion control system based on machine learning according to claim 1, characterized in that: Feature extraction is performed on historical combustion data from several historical combustion logs to obtain combustion state data for each operating condition category, including: Obtain historical combustion logs for all operating conditions and perform cluster analysis to obtain multiple operating condition categories and several historical combustion logs for each operating condition category; Preset a number of combustion evaluation indicators, and set each combustion evaluation indicator to strongly correlate historical combustion data in a number of historical combustion logs for each operating condition category; A time reference line is established based on the historical monitoring duration of each historical combustion log of each operating condition category, and a collection node is established based on a preset time interval. Multiple sampling intensity-related historical combustion data and other historical combustion data in the corresponding historical combustion log are collected according to the collection node; Generate an initial combustion state assessment value of the corresponding combustion evaluation index based on all strongly correlated historical combustion data at the same acquisition node; Mapping the initial combustion state evaluation values of all combustion evaluation indicators at multiple acquisition nodes and multiple historical combustion data onto the time reference line of the corresponding historical combustion log to obtain multiple initial combustion state evaluation value change curves and multiple historical combustion data change curves; Perform dependency analysis on each historical combustion data change curve in the same historical combustion log and a number of initial combustion state assessment value change curves to obtain the dependency degree; A correlation coefficient is set between each historical combustion data change curve and the corresponding initial combustion state evaluation value change curve according to the degree of dependence. If the correlation coefficient is greater than a preset correlation coefficient threshold, the corresponding historical combustion data is set as the associated historical combustion data of the corresponding combustion evaluation index. Comparing and analyzing the associated historical combustion data of the same combustion evaluation index determined in each historical combustion log of the same operating condition category to obtain the occurrence ratio of the same associated historical combustion data, and eliminating the associated historical combustion data with an occurrence ratio less than a preset ratio threshold; The strongly correlated historical combustion data of each combustion evaluation index of the same operating condition category and the remaining correlated historical combustion data are set as the combustion state data of the corresponding operating condition category, and a weight coefficient of each combustion state data is set.

3. The multi-condition coal combustion control system based on machine learning according to claim 2, characterized in that: A combustion state assessment model is constructed based on the combustion state data of each operating condition category, including: Obtaining combustion state data for each operating condition category at a preset historical time node, and generating combustion state evaluation values for a number of combustion evaluation indicators at the corresponding preset historical time nodes; Generate a comprehensive combustion state evaluation value at a corresponding historical time node based on the combustion state evaluation values of several combustion evaluation indicators at the same preset historical time node; The combustion state data at multiple preset historical time nodes of each operating condition category are used as training input data, and the comprehensive combustion state evaluation values at the corresponding preset historical time nodes are used as training output data. Neural network training is performed to obtain a combustion state evaluation model.

4. The multi-condition coal combustion control system based on machine learning according to claim 3, characterized in that: The combustion control model includes: Divide multiple preset historical time nodes into several analysis periods according to preset time intervals; Analyze the comprehensive combustion state evaluation value of each operating condition category in each analysis period, and determine the change value and change trend of the comprehensive combustion state evaluation value in the corresponding analysis period; If the change trend is normal and the change value is greater than the preset change value threshold, the period in the analysis period that is in the normal trend is intercepted and set as the focus period; If the final comprehensive combustion state evaluation value of the focus period is greater than the preset evaluation value threshold, a number of historical control parameters in the corresponding focus period are retrieved; Obtaining historical control adjustment values of the retrieved historical control parameters in the corresponding focus period, and constructing a historical control parameter-adjustment value mapping table for the corresponding focus period; Mapping the historical control parameter-adjustment value mapping table again with the combustion state data at the first preset historical time node of the corresponding focus period to obtain a combustion state data-historical control parameter-adjustment value mapping table; The combustion state data in a plurality of combustion state data-historical control parameter-control value mapping tables are used as training input data, and the historical control parameter-control value mapping tables corresponding to the combustion state data are used as training output data to perform neural network training and obtain a combustion control model.

5. The multi-condition coal combustion control system based on machine learning according to claim 4, characterized in that: Determine whether optimization control is needed based on the real-time comprehensive combustion state evaluation value, including: Determine the real-time operating condition category and obtain the real-time combustion state data at the current monitoring time node, input the real-time operating condition category and the real-time combustion state data into the combustion state evaluation model to obtain a real-time comprehensive combustion state evaluation value; Constructing a real-time combustion state data change curve according to the real-time combustion state data at multiple monitoring time nodes in the current monitoring cycle; Preset several monitoring time nodes in the future period, perform curve trend extrapolation on the real-time combustion state data change curve, obtain predicted combustion state data at several monitoring time nodes in the future period, and input the data into the combustion state assessment model to obtain predicted comprehensive combustion state assessment values at several monitoring time nodes in the future period; generating a compensation coefficient for the real-time comprehensive combustion state evaluation value at the current monitoring time node according to the change characteristics of the predicted comprehensive combustion state evaluation values at several monitoring time nodes in the future period; The real-time comprehensive combustion state evaluation value is corrected according to the compensation coefficient. If the corrected real-time comprehensive combustion state evaluation value is less than the preset evaluation value, it is judged that optimization control is required. If the corrected real-time comprehensive combustion state evaluation value is greater than the preset evaluation value, it is judged that optimization control is not required.

6. The multi-condition coal combustion control system based on machine learning according to claim 5, characterized in that: Generate a first combustion control strategy based on the combustion control model and issue a control instruction, including: If optimization control is required, the real-time combustion state data at the current monitoring time node is input into the combustion control model to obtain the control values of several control parameters and generate a first combustion control strategy; Generate a standard change rate and a standard change value for each combustion state data after the first combustion control strategy, generate a standard change model for the corresponding combustion state data according to the standard change rate and the standard change value, and set a number of preset feedback time nodes; A control instruction is generated according to the first combustion control strategy, and a monitoring instruction is generated according to a preset feedback time node.

7. The multi-condition coal combustion control system based on machine learning according to claim 6, characterized in that: Obtain real-time combustion status data based on preset feedback time nodes and determine whether to make corrections, including: Acquire real-time combustion state data based on preset feedback time nodes, and calculate the actual change rate and actual change value at the corresponding preset feedback time nodes; The actual change rate and actual change value of each real-time combustion state data at the preset feedback time node are input into the standard change model to obtain the difference degree; Generate a comprehensive difference degree of the corresponding preset feedback time node according to the difference degree of each real-time combustion state data at the same preset feedback time node, and construct a comprehensive difference degree sequence according to the time sequence of the preset feedback time nodes; If the comprehensive difference degree sequence does not show a downward trend, it is determined that the first combustion control strategy needs to be modified.

8. The multi-condition coal combustion control system based on machine learning according to claim 7, characterized in that: The preset correction reference library includes: A plurality of preset differences are set according to a standard variation model of each real-time combustion state data, and each preset difference is mapped to a plurality of preset control parameters to be corrected and corresponding preset correction values; Randomly combining multiple preset difference degrees of different real-time combustion state data to obtain multiple preset comprehensive difference degrees; Perform conflict analysis based on a number of preset control parameters to be modified and their corresponding preset correction values mapped to the same preset comprehensive difference, and determine whether to perform adjustment and optimization based on the analysis results; If so, generating a preset correction strategy corresponding to the preset comprehensive difference according to the optimized plurality of preset control parameters to be corrected and the corresponding preset correction values; A preset correction reference library is constructed based on multiple preset comprehensive differences and corresponding preset correction strategies.

9. The multi-condition coal combustion control system based on machine learning according to claim 8, characterized in that: Generates correction instructions based on a preset correction reference library, including: Obtaining the comprehensive difference at the current preset feedback time node, and performing similarity analysis with the preset comprehensive difference to obtain multiple similarities; The preset correction strategy mapped to the preset comprehensive difference degree with the greatest similarity is set as the correction strategy of the first combustion control strategy, and a correction instruction is generated.

10. A multi-condition coal combustion control method based on machine learning, characterized in that: include: Determine multiple operating condition categories and obtain several historical combustion logs for each operating condition category, perform feature extraction on the historical combustion data in the several historical combustion logs, and obtain combustion state data for each operating condition category; Construct a combustion state evaluation model and a combustion control model based on the combustion state data of each operating condition category, and generate a real-time comprehensive combustion state evaluation value based on the combustion state evaluation model; Determine whether optimization control is required based on the real-time comprehensive combustion state evaluation value, and if so, generate a first combustion control strategy based on the combustion control model and issue a control instruction; The real-time combustion status data is obtained based on the preset feedback time node, and it is determined whether correction is required. If so, a correction instruction is generated based on the preset correction reference library.

Images