Method for evaluating state of steam turbine blade

Abstract

The invention relates to the technical field of steam turbine blade state evaluation, and discloses a steam turbine blade state evaluation method, which comprises the following steps: dividing a historical monitoring log according to a state type to obtain a plurality of monitoring log sets; first historical monitoring data and second historical monitoring data of historical monitoring logs in each monitoring log set are extracted and analyzed, and attention factors, correlation factors of the attention factors and corresponding weight coefficients are determined; constructing a state evaluation model according to the attention factors and the corresponding weight coefficients, and constructing an auxiliary evaluation model according to the correlation factors of the attention factors and the corresponding weight coefficients; generating a prediction state coefficient according to the state evaluation model, and obtaining a correction coefficient according to the auxiliary evaluation model; whether an alarm instruction of the turbine blade is generated or not is judged according to the corrected prediction state coefficient, the actual state of the turbine blade can be accurately evaluated, and normal and stable operation of a turbine is guaranteed.

Description

Technical Field

[0001] This application relates to the technical field of steam turbine blade condition assessment, and particularly to a method for assessing the condition of steam turbine blades. Background Art

[0002] Steam turbines are one of the core equipment in thermal power and nuclear power enterprises. Their operating conditions directly affect the economic benefits of the entire enterprise. However, for a long time, the safety assessment of steam turbines has mainly focused on the rotor system, and little attention has been paid to the safety assessment of blades, greatly reducing the accuracy of steam turbine condition monitoring, resulting in blade damage during the operation of steam turbines and affecting the normal operation of steam turbines.

[0003] In the prior art, the condition assessment of steam turbine blades mostly relies on manual inspection after shutdown. During the operation of steam turbines, the actual conditions of steam turbine blades cannot be accurately reflected, reducing the efficiency of steam turbine blade condition assessment and unable to ensure the normal and stable operation of steam turbines. Summary of the Invention

[0004] To solve the above technical problems, this application provides a method for assessing the condition of steam turbine blades. By analyzing the first historical monitoring data and the second historical monitoring data in the monitoring log sets of different state types, the attention factors, the associated factors of the attention factors, and the corresponding weight coefficients are determined, and a condition assessment model and an auxiliary assessment model are constructed. According to the condition assessment model, the predicted condition coefficient is determined, and according to the auxiliary assessment model, the correction coefficient is determined. Whether an alarm instruction occurs is judged based on the correction coefficient and the predicted condition coefficient, and the actual condition of the steam turbine blades can be accurately assessed to ensure the normal and stable operation of the steam turbines.

[0005] In some embodiments of this application, a method for assessing the condition of steam turbine blades is provided, including: Obtain a number of historical monitoring logs of steam turbine blades, analyze the historical monitoring logs, determine the state types of the steam turbine blades in the historical monitoring logs, and divide the historical monitoring logs according to the state types to obtain multiple monitoring log sets, where the state types include normal state, abnormal state, and suspected abnormal state; Extract the first historical monitoring data and the second historical monitoring data of the historical monitoring logs in each monitoring log set, analyze the first historical monitoring data to determine the attention factors and the corresponding weight coefficients, and perform correlation analysis on the attention factors and the second historical monitoring data to determine the associated factors of the attention factors and the corresponding weight coefficients; Construct a condition assessment model according to the attention factors and the corresponding weight coefficients, and construct an auxiliary assessment model according to the associated factors of the attention factors and the corresponding weight coefficients; Generate a predicted state coefficient of the steam turbine blade at the current monitoring time node according to the state evaluation model, and obtain a correction coefficient according to the auxiliary evaluation model; Correct the predicted state coefficient according to the correction coefficient, and determine whether to generate an alarm instruction for the steam turbine blade according to the corrected predicted state coefficient.

[0006] In some embodiments of the present application, divide the historical monitoring logs according to the state type to obtain a plurality of monitoring log sets, including: Obtain the historical state coefficient of the steam turbine blade in each historical monitoring log, compare the historical state coefficient with the standard state coefficient interval when the steam turbine blade is operating normally, and determine the state type of the steam turbine blade in each historical monitoring log according to the comparison result; If the historical state coefficient is within the standard state coefficient interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the normal state; If the historical state coefficient is not within the standard state coefficient interval, calculate the difference between the historical state coefficient and the historical state coefficient in the standard state coefficient interval; When the difference between the historical state coefficients is within the first preset difference interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the suspected abnormal state; If the difference between the historical state coefficients is within the second preset difference interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the abnormal state; Wherein, the first preset difference interval is smaller than the second preset difference interval; Divide a number of historical monitoring logs according to the state type of the historical monitoring logs to obtain a plurality of monitoring log sets, and the monitoring log sets are respectively the monitoring log set in the normal state, the monitoring log set in the abnormal state, and the monitoring log set in the suspected abnormal state.

[0007] In some embodiments of the present application, analyze the first historical monitoring data to determine the attention factors and the corresponding weight coefficients, including: Determine the normal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data in different historical monitoring logs in the monitoring log set in the normal state; Determine the abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data in different historical monitoring logs in the monitoring log set in the abnormal state; Determine the suspected abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data in different historical monitoring logs in the monitoring log set in the suspected abnormal state; Compare the normal monitoring data intervals of each first historical monitoring data with the corresponding abnormal monitoring data intervals and suspected abnormal monitoring data intervals respectively to determine the first data difference and the second data difference of the corresponding first historical monitoring data; If the first data difference of the first historical monitoring data is greater than the second data difference and the second data difference is greater than the preset data difference threshold, set the corresponding first historical monitoring data as a factor to be determined for attention; Calculate the first undetermined data difference between the first historical feature data of each historical monitoring log in the monitoring log set in the normal state and the second historical feature data of each historical monitoring log in the monitoring log set in the suspected abnormal state for each factor to be determined for attention, and the second undetermined data difference between the third historical feature data of each historical monitoring log in the monitoring log set in the abnormal state; Set the accuracy of the factor to be determined for attention according to the first quantity of historical monitoring logs with the first undetermined data difference greater than the preset data difference threshold and the second quantity of historical monitoring data with the second undetermined data difference greater than the first undetermined data difference; If the accuracy is greater than the preset accuracy threshold, set the factor to be determined for attention as a factor for attention; Generate the weight coefficient of the corresponding factor for attention according to the first attention difference of multiple first undetermined data differences of each factor for attention greater than the preset data difference threshold, the second attention difference of multiple second undetermined data differences greater than the first undetermined data difference, and the accuracy;

[0008] In some embodiments of the present application, the calculation formula for the accuracy of the factor to be determined for attention is: ; Where Z is the accuracy, is the first accurate conversion coefficient, is the second accurate conversion coefficient, d1 is the first quantity of historical monitoring logs with the first undetermined data difference greater than the preset data difference threshold, m2 is the total number of historical monitoring logs in the monitoring log set in the suspected abnormal state, d2 is the second quantity of historical monitoring data with the second undetermined data difference greater than the first undetermined data difference, and m3 is the total number of historical monitoring logs in the monitoring log set in the abnormal state; The calculation formula for the weight coefficient of the factor for attention is: ; Where, Q1 is the weight coefficient of the factor for attention, q1 is the weight conversion coefficient, is the i1th first attention difference, a1 is the weight coefficient of the first attention difference, is the i2th second attention difference, and a2 is the weight coefficient of the second attention difference.

[0009] In some embodiments of the present application, determining the associated factors of the attention factors and the corresponding weight coefficients includes: Taking the time length of the historical monitoring period of each historical monitoring log as the time reference line, setting a plurality of data acquisition nodes according to a preset time interval, obtaining a plurality of attention factors and second historical monitoring data in the corresponding historical monitoring log according to the data acquisition nodes, and mapping them onto the time reference line to obtain an analysis diagram of the corresponding historical monitoring log; Among them, the analysis diagram includes the data change curve of each attention factor in the corresponding historical monitoring log and the data change curve of each second historical monitoring data; Mark the mutation time nodes of the data change curve of each attention factor in the analysis diagram, and intercept the first curve segment in the adjacent time period before the marked mutation time node of the data change curve of each second historical monitoring data for each attention factor. There is at least one marked mutation time node for each attention factor; Map the first curve segment of each second historical monitoring data to the marked mutation time node of each attention factor to obtain a first curve segment - mutation time node mapping table of each second historical monitoring data for each attention factor; Calculate the degree of fluctuation of each second historical monitoring data in the first curve segment; If the degree of fluctuation of all the first curve segments in the first curve segment - mutation time node mapping table of the same second historical monitoring data for the same attention factor is greater than the preset fluctuation degree threshold, then set the second historical monitoring data as the associated factor of the corresponding attention factor.

[0010] In some embodiments of the present application, determining the associated factors of the attention factors and the corresponding weight coefficients includes: Obtain the associated factors of each attention factor, and the mapping relationship between the degree of mutation at the marked mutation time node of each attention factor and the degree of fluctuation of the associated factors in the monitoring log set of different state types. Quantify the degree of mutation and the degree of fluctuation with mapping relationships to obtain a mutation quantification value and the corresponding fluctuation quantification value; Construct a first fluctuation quantification value - mutation quantification value mapping table, a second fluctuation quantification value - mutation quantification value mapping table, and a third fluctuation quantification value - mutation quantification value mapping table respectively according to the mutation quantification values and the corresponding fluctuation quantification values of each attention factor and the corresponding associated factors in different state types; Determine the first influence coefficient of each associated factor for the corresponding attention factor in the normal state according to the first fluctuation quantification value - mutation quantification value mapping table; Determine the second influence coefficient of each associated factor for the corresponding attention factor in the suspected abnormal state according to the second fluctuation quantification value - mutation quantification value mapping table; Determine the third influence coefficient of each associated factor on the corresponding concerned factor in the abnormal state according to the third fluctuation quantization value - mutation quantization value mapping table; Determine the comprehensive influence coefficient of each associated factor on the corresponding concerned factor according to the first influence coefficient, the second influence coefficient and the third influence coefficient; Generate the weight coefficient of the corresponding associated factor according to the number of concerned factors corresponding to the same associated factor and the comprehensive influence coefficient of the associated factor on each corresponding concerned factor; The calculation formula of the weight coefficient of the said associated factor is: ; Wherein, Q2 is the weight coefficient of the associated factor, q2 is the weight conversion coefficient of the associated factor, n1 is the number of concerned factors corresponding to the current associated factor, n2 is the total number of concerned factors, and Ys is the comprehensive influence coefficient of the associated factor on the s-th concerned factor.

[0011] In some embodiments of the present application, a state evaluation model is constructed according to the concerned factors and the corresponding weight coefficients, and an auxiliary evaluation model is constructed according to the associated factors of the concerned factors and the corresponding weight coefficients, including: Using the concerned factors and the corresponding weight coefficients in the historical monitoring logs of different state types as training input data, and the historical state coefficients corresponding to the historical monitoring logs as training output data, perform neural network training to obtain the state evaluation model of the steam turbine blade; Using the associated factors of each concerned factor, the fluctuation degree of the associated factor and the weight coefficient of the associated factor as training input data, and the mutation degree of the corresponding concerned factor as training output data, perform neural network training to obtain the auxiliary evaluation model of the steam turbine blade.

[0012] In some embodiments of the present application, a predicted state coefficient of the steam turbine blade at the current monitoring time node is generated according to the state evaluation model, and a correction coefficient is obtained according to the auxiliary evaluation model, including: Obtain the real-time concerned factors at the current monitoring time node, input the real-time concerned factors and the corresponding weight coefficients into the state evaluation model to obtain the predicted state coefficient; Obtain the historical associated factors of each concerned factor and the fluctuation degree of the historical associated factors in the previous preset time period, and combine the weight coefficients of the historical associated factors, and input them into the auxiliary evaluation model to obtain the predicted mutation degree of the corresponding concerned factor; Generate the predicted concerned factors at the current monitoring time node according to the historical concerned factors and the predicted mutation degree in the previous preset time period, compare the predicted concerned factors with the corresponding real-time concerned factors to obtain the difference value; Generate a deviation degree based on multiple difference magnitudes and the weight coefficients of corresponding attention factors. Set a correction coefficient for the prediction status coefficient according to the deviation degree.

[0013] In some embodiments of the present application, determining whether to generate an alarm instruction for a steam turbine blade according to the corrected prediction status coefficient includes: Obtain the historical status coefficient of the previous monitoring time node, and determine the change characteristics of the status coefficient according to the prediction status coefficient of the current monitoring time node and the historical status coefficient of the previous monitoring time node. The change characteristics include the change trend, change rate, and change magnitude. Generate a compensation coefficient according to the change trend, change rate, and change magnitude of the status coefficient. Preset a status coefficient threshold and a compensation coefficient threshold in advance. When the corrected prediction status coefficient is less than the status coefficient threshold, send an alarm instruction. When the corrected prediction status coefficient is not less than the status coefficient threshold and the compensation coefficient is less than the compensation coefficient threshold, send a warning signal. When the corrected prediction status coefficient is not less than the status coefficient threshold and the compensation coefficient is greater than the compensation coefficient threshold, do not send an alarm instruction and a warning instruction.

[0014] A method for evaluating the status of a steam turbine blade according to an embodiment of the present application, compared with the prior art, has the beneficial effect that: By analyzing the first historical monitoring data and the second historical monitoring data in the monitoring log sets of different status types, determine the attention factors, associated factors of the attention factors, and corresponding weight coefficients, and construct a status evaluation model and an auxiliary evaluation model. Determine the prediction status coefficient according to the status evaluation model, determine the correction coefficient according to the auxiliary evaluation model, and determine whether an alarm instruction occurs according to the correction coefficient and the prediction status coefficient, which can accurately evaluate the actual status of the steam turbine blade and ensure the normal and stable operation of the steam turbine. Brief Description of the Drawings

[0015] Figure 1 It is a schematic flowchart of a method for evaluating the status of a steam turbine blade according to an embodiment of the present application. Detailed Embodiments

[0016] The following combines the drawings and embodiments to further describe in detail the specific embodiments of the present application. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.

[0017] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present application.

[0018] The terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0019] In the description of the present application, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0020] As Figure 1 shown, a method for evaluating the state of a steam turbine blade according to an embodiment of the present application includes: Step S101: Obtain a number of historical monitoring logs of the steam turbine blade, analyze the historical monitoring logs, determine the state types of the steam turbine blade in the historical monitoring logs, and divide the historical monitoring logs according to the state types to obtain a plurality of monitoring log sets, where the state types include normal state, abnormal state, and suspected abnormal state; Step S102: Extract the first historical monitoring data and the second historical monitoring data of the historical monitoring logs in each monitoring log set, analyze the first historical monitoring data to determine the attention factors and the corresponding weight coefficients, and perform correlation analysis on the attention factors and the second historical monitoring data to determine the correlation factors of the attention factors and the corresponding weight coefficients; Step S103: Construct a state evaluation model according to the attention factors and the corresponding weight coefficients, and construct an auxiliary evaluation model according to the correlation factors of the attention factors and the corresponding weight coefficients; Step S104: Generate a predicted state coefficient of the steam turbine blade at the current monitoring time node according to the state evaluation model, and obtain a correction coefficient according to the auxiliary evaluation model; Step S105: Correct the predicted state coefficient according to the correction coefficient, and determine whether to generate an alarm instruction for the steam turbine blade according to the corrected predicted state coefficient.

[0021] In this embodiment, the first historical monitoring data refers to the directly related data of the steam turbine blade in the corresponding historical monitoring log. For example, the crack condition of the steam turbine blade, the blade vibration frequency, the blade stress condition, etc. The second historical monitoring data refers to other operation data in the corresponding historical monitoring log.

[0022] In this embodiment, the associated factor of the attention factor refers to the historical operation data that causes the change of the attention factor, that is, the attention factor changes with the change of the historical operation data.

[0023] In this embodiment, by determining the attention factor and the corresponding weight coefficient, the accuracy of the state evaluation of the steam turbine blade is improved. By determining the associated factor of the attention factor and the corresponding weight coefficient, the credibility of the predicted state coefficient is accurately judged, ensuring the normal and stable operation of the steam turbine.

[0024] In some embodiments of the present application, the historical monitoring log is divided according to the state type to obtain a plurality of monitoring log sets, including: Obtain the historical state coefficient of the steam turbine blade in each historical monitoring log, compare the historical state coefficient with the standard state coefficient interval when the steam turbine blade operates normally, and determine the state type of the steam turbine blade in each historical monitoring log according to the comparison result; If the historical state coefficient is within the standard state coefficient interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the normal state; If the historical state coefficient is not within the standard state coefficient interval, calculate the difference between the historical state coefficient and the historical state coefficient in the standard state coefficient interval; When the difference between the historical state coefficients is within the first preset difference interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the suspected abnormal state; If the difference between the historical state coefficients is within the second preset difference interval, set the state type of the steam turbine blade in the corresponding historical monitoring log to the abnormal state; Wherein, the first preset difference interval is smaller than the second preset difference interval; Divide a number of historical monitoring logs according to the state type of the historical monitoring log to obtain a plurality of monitoring log sets, and the monitoring log sets are respectively the monitoring log set in the normal state, the monitoring log set in the abnormal state, and the monitoring log set in the suspected abnormal state.

[0025] In some embodiments of the present application, analyzing the first historical monitoring data to determine the attention factor and the corresponding weight coefficient includes: Determine the normal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set of the normal state; Determine the abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set of the abnormal state; Determine the suspected abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set of the suspected abnormal state; Compare the normal monitoring data interval of each first historical monitoring data with the corresponding abnormal monitoring data interval and suspected abnormal monitoring data interval respectively to determine the first data difference and the second data difference of the corresponding first historical monitoring data; If the first data difference of the first historical monitoring data is greater than the second data difference and the second data difference is greater than the preset data difference threshold, set the corresponding first historical monitoring data as a factor to be determined for attention; Calculate the first undetermined data difference between the first historical feature data of each historical monitoring log in the monitoring log set of the normal state and the second historical feature data of each historical monitoring log in the monitoring log set of the suspected abnormal state for each factor to be determined for attention, and the second undetermined data difference between the third historical feature data of each historical monitoring log in the monitoring log set of the abnormal state; Set the accuracy of the factor to be determined for attention according to the first quantity of historical monitoring logs with the first undetermined data difference greater than the preset data difference threshold and the second quantity of historical monitoring data with the second undetermined data difference greater than the first undetermined data difference; If the accuracy is greater than the preset accuracy threshold, set the factor to be determined for attention as a factor for attention; Generate the weight coefficient corresponding to each factor for attention according to the first attention difference with multiple first undetermined data differences greater than the preset data difference threshold, the second attention difference with multiple second undetermined data differences greater than the first undetermined data difference, and the accuracy.

[0026] In this embodiment, the first historical feature data is obtained by performing mean processing on the corresponding first historical monitoring data whose stable duration in each historical monitoring log in the monitoring log set of each to-be-determined attention factor in the normal state is greater than a preset duration. The second historical feature data is obtained by performing mean processing on the corresponding first historical monitoring data whose stable duration in each historical monitoring log in the monitoring log set of each to-be-determined attention factor in the suspected abnormal state is greater than a preset duration, and the above-mentioned stable duration appears after a large change node in the first historical monitoring data. The third historical feature data is obtained by performing mean processing on the corresponding first historical monitoring data whose stable duration in each historical monitoring log in the monitoring log set of each to-be-determined attention factor in the abnormal state is greater than a preset duration, and the above-mentioned stable duration appears after a large change node in the first historical monitoring data. Wherein, the stable duration refers to a period when the data fluctuation degree of the first historical monitoring data is small.

[0027] In some embodiments of the present application, the calculation formula for the accuracy of the to-be-determined attention factor is: ; Where Z is the accuracy, is the first accurate conversion coefficient, is the second accurate conversion coefficient, d1 is the first quantity of historical monitoring logs whose first to-be-determined data difference is greater than the preset data difference threshold, m2 is the total number of historical monitoring logs in the monitoring log set in the suspected abnormal state, d2 is the second quantity of historical monitoring data whose second to-be-determined data difference is greater than the first to-be-determined data difference, and m3 is the total number of historical monitoring logs in the monitoring log set in the abnormal state; The calculation formula for the weight coefficient of the attention factor is: ; Where Q1 is the weight coefficient of the attention factor, q1 is the weight conversion coefficient, is the i1th first attention difference, a1 is the weight coefficient of the first attention difference, is the i2th second attention difference, and a2 is the weight coefficient of the second attention difference.

[0028] In this embodiment, when d1 / m2 is larger and d2 / m3 is larger, the corresponding first accurate conversion coefficient and second accurate conversion coefficient are larger, that is, the corresponding accuracy is larger, and the value range of the accuracy is (0, 1).

[0029] In this embodiment, when the sum of multiple first attention differences is smaller and the sum of multiple second attention differences is smaller, the corresponding weight conversion coefficient is larger, that is, the weight coefficient of the corresponding attention factor is larger.

[0030] In this embodiment, when the change degree of the first attention difference and the second attention difference is smaller, it indicates that the data change of the corresponding attention factor is smaller, but it still causes a change in the state coefficient of the steam turbine blade. That is, the corresponding attention factor is more important for the state evaluation of the steam turbine blade, that is, the weight coefficient of the corresponding attention factor is larger.

[0031] In this embodiment, by determining the attention factors and the corresponding weight coefficients of the steam turbine blade, it lays a foundation for establishing a state evaluation model later and improves the accuracy of the state evaluation of the steam turbine blade.

[0032] In some embodiments of the present application, determining the associated factors of the attention factors and the corresponding weight coefficients includes: Taking the time length of the historical monitoring period of each historical monitoring log as the time reference line, setting a plurality of data acquisition nodes according to a preset time interval, obtaining a plurality of attention factors and second historical monitoring data in the corresponding historical monitoring log according to the data acquisition nodes, and mapping them to the time reference line to obtain an analysis diagram of the corresponding historical monitoring log; Wherein, the analysis diagram includes the data change curve of each attention factor in the corresponding historical monitoring log and the data change curve of each second historical monitoring data; Marking the mutation time nodes of the data change curves of each attention factor in the analysis diagram, and intercepting the first curve segments in the adjacent time periods before the marked mutation time nodes of each attention factor of the data change curves of each second historical monitoring data. There is at least one marked mutation time node for each attention factor; Mapping the first curve segments of each second historical monitoring data to the marked mutation time nodes of each attention factor to obtain a first curve segment - mutation time node mapping table of each second historical monitoring data for each attention factor; Calculating the fluctuation degree of each second historical monitoring data in the first curve segment; If the fluctuation degrees of all the first curve segments in the first curve segment - mutation time node mapping table of the same second historical monitoring data for the same attention factor are greater than the preset fluctuation degree threshold, then the second historical monitoring data is set as the associated factor of the corresponding attention factor.

[0033] In this embodiment, the mutation time node refers to the time node with a large change degree in the data change curve of each attention factor.

[0034] In this embodiment, by determining the mutation time nodes of each concerned factor, intercepting a plurality of first curve segments in the adjacent time period before the mutation time node, judging the correlation between each concerned factor and the second historical monitoring data according to the fluctuation degree of the first curve segments, and selecting the correlation factors of each concerned factor, a foundation is laid for subsequent construction of an auxiliary evaluation model and improvement of the accuracy of predicting the state coefficient.

[0035] In some embodiments of the present application, determining the correlation factors and corresponding weight coefficients of the concerned factors includes: Obtaining the correlation factors of each concerned factor, and the mapping relationship between the mutation degree at the marked mutation time node of each concerned factor and the fluctuation degree of the correlation factors in the monitoring log set of different state types, quantifying the existing mutation degree and fluctuation degree to obtain a mutation quantification value and a corresponding fluctuation quantification value; Constructing a first fluctuation quantification value - mutation quantification value mapping table, a second fluctuation quantification value - mutation quantification value mapping table, and a third fluctuation quantification value - mutation quantification value mapping table respectively according to the mutation quantification values and corresponding fluctuation quantification values of each concerned factor and its corresponding correlation factor in different state types; Determining the first influence coefficient of each correlation factor on the corresponding concerned factor under the normal state according to the first fluctuation quantification value - mutation quantification value mapping table; Determining the second influence coefficient of each correlation factor on the corresponding concerned factor under the suspected abnormal state according to the second fluctuation quantification value - mutation quantification value mapping table; Determining the third influence coefficient of each correlation factor on the corresponding concerned factor under the abnormal state according to the third fluctuation quantification value - mutation quantification value mapping table; Determining the comprehensive influence coefficient of each correlation factor on the corresponding concerned factor according to the first influence coefficient, the second influence coefficient, and the third influence coefficient; Generating the weight coefficient of the corresponding correlation factor according to the number of concerned factors corresponding to the same correlation factor and the comprehensive influence coefficient of the correlation factor on each corresponding concerned factor; The calculation formula for the weight coefficient of the correlation factor is: ; where Q2 is the weight coefficient of the correlation factor, q2 is the weight conversion coefficient of the correlation factor, n1 is the number of concerned factors corresponding to the current correlation factor, n2 is the total number of concerned factors, and Ys is the comprehensive influence coefficient of the correlation factor on the sth concerned factor.

[0036] In this embodiment, the first influence coefficient is obtained by taking the ratio of multiple fluctuation quantization values and the corresponding mutation quantization values in the first fluctuation quantization value - mutation quantization value mapping table. The second influence coefficient is obtained by taking the ratio of multiple fluctuation quantization values and the corresponding mutation quantization values in the second fluctuation quantization value - mutation quantization value mapping table. The third influence coefficient is obtained by taking the ratio of multiple fluctuation quantization values and the corresponding mutation quantization values in the third fluctuation quantization value - mutation quantization value mapping table.

[0037] In this embodiment, the comprehensive influence coefficient is calculated based on the first influence coefficient, the second influence coefficient, and the third influence coefficient. When the differences among the first influence coefficient, the second influence coefficient, and the third influence coefficient are not significant, the mean value is processed to obtain the comprehensive influence coefficient. When the differences among the first influence coefficient, the second influence coefficient, and the third influence coefficient are large, the outliers are removed and the mean value is calculated to obtain the comprehensive influence coefficient.

[0038] In this embodiment, by calculating the correlation factors of the concerned factors and calculating the weight coefficients of each correlation factor, it lays a foundation for constructing an auxiliary evaluation model in the subsequent stage, improves the accuracy and reliability of predicting the state coefficient, and ensures the accurate evaluation of the state of the steam turbine blade and the normal and stable operation of the steam turbine.

[0039] In some embodiments of the present application, a state evaluation model is constructed according to the concerned factors and the corresponding weight coefficients, and an auxiliary evaluation model is constructed according to the correlation factors of the concerned factors and the corresponding weight coefficients, including: Taking the concerned factors and the corresponding weight coefficients in the historical monitoring logs of different state types as training input data, and the historical state coefficients corresponding to the historical monitoring logs as training output data, and performing neural network training to obtain the state evaluation model of the steam turbine blade; Taking the correlation factors of each concerned factor, the fluctuation degree of the correlation factor, and the weight coefficient of the correlation factor as training input data, and the mutation degree corresponding to the concerned factor as training output data, and performing neural network training to obtain the auxiliary evaluation model of the steam turbine blade.

[0040] In some embodiments of the present application, a predicted state coefficient of the steam turbine blade at the current monitoring time node is generated according to the state evaluation model, and a correction coefficient is obtained according to the auxiliary evaluation model, including: Obtaining the real-time concerned factors at the current monitoring time node, and inputting the real-time concerned factors and the corresponding weight coefficients into the state evaluation model to obtain the predicted state coefficient; Obtaining the historical correlation factors of each concerned factor and the fluctuation degree of the historical correlation factors in the previous preset period, and combining the weight coefficients of the historical correlation factors, and inputting them into the auxiliary evaluation model to obtain the predicted mutation degree of the corresponding concerned factor; Generate the predicted attention factor at the current monitoring time node based on the historical attention factor in the previous preset time period and the predicted mutation degree, compare the predicted attention factor with the corresponding real-time attention factor, and obtain the difference value; Generate the deviation degree according to multiple difference values and the weight coefficients of the corresponding attention factors; Set the correction coefficient of the predicted state coefficient according to the deviation degree.

[0041] In this embodiment, when the difference value is larger and the corresponding weight coefficient is larger, the deviation degree is larger. When the deviation degree is larger, the corresponding correction coefficient is smaller, and vice versa. The value range of the correction coefficient is (0, 1).

[0042] In this embodiment, determine the predicted attention factor at the monitoring time node according to the auxiliary evaluation model, prevent the acquisition error of the attention factor, improve the credibility and accuracy of the predicted state coefficient, and thus ensure the normal and stable operation of the steam turbine.

[0043] In some embodiments of the present application, determine whether to generate an alarm instruction for the steam turbine blade according to the corrected predicted state coefficient, including: Obtain the historical state coefficient of the previous monitoring time node, and determine the change characteristics of the state coefficient according to the predicted state coefficient of the current monitoring time node and the historical state coefficient of the previous monitoring time node. The change characteristics include the change trend, change rate, and change value; Generate a compensation coefficient according to the change trend, change rate, and change value of the state coefficient; Preset the state coefficient threshold and the compensation coefficient threshold; When the corrected predicted state coefficient is less than the state coefficient threshold, send an alarm instruction; When the corrected predicted state coefficient is not less than the state coefficient threshold and the compensation coefficient is less than the compensation coefficient threshold, send a warning signal; When the corrected predicted state coefficient is not less than the state coefficient threshold and the compensation coefficient is greater than the compensation coefficient threshold, do not send an alarm instruction and a warning instruction.

[0044] In this embodiment, the corrected predicted state coefficient = predicted state coefficient * correction coefficient, and the state coefficient threshold is the minimum state coefficient of the blade when the steam turbine operates normally.

[0045] In this embodiment, the compensation coefficient threshold is determined according to the critical normal change characteristics of the state coefficient. The compensation coefficient = B0 * △B1 * (A1 * B2 + B3 * A2), where △B1 is the selection coefficient of the change trend. When the change trend is an upward trend, the selection coefficient is 1. When the change trend is a downward trend, the selection coefficient is 1. B2 is the change amount value, B3 is the change rate, and B0 is the compensation conversion coefficient. When the change trend is an upward trend, the larger the change amount value and the change rate, the larger the corresponding compensation coefficient, and vice versa.

[0046] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art of the present technology, without departing from the technical principle of the present application, several improvements and replacements can be made, and these improvements and replacements should also be regarded as the protection scope of the present application.

Claims

1. A method for evaluating the condition of a steam turbine blade, characterized in that: include: Acquire several historical monitoring logs of the steam turbine blades, analyze the historical monitoring logs, determine the state types of the steam turbine blades in the historical monitoring logs, and divide the historical monitoring logs according to the state types to obtain multiple monitoring log sets, wherein the state types include normal state, abnormal state, and suspected abnormal state; Extracting first historical monitoring data and second historical monitoring data of the historical monitoring log in each monitoring log set, analyzing the first historical monitoring data to determine a concern factor and a corresponding weight coefficient, and performing correlation analysis on the concern factor and the second historical monitoring data to determine a correlation factor of the concern factor and a corresponding weight coefficient; A state assessment model is constructed based on the focus factors and the corresponding weight coefficients, and an auxiliary assessment model is constructed based on the correlation factors of the focus factors and the corresponding weight coefficients; Generate a predicted state coefficient of the steam turbine blade at the current monitoring time node according to the state assessment model, and obtain a correction coefficient according to the auxiliary assessment model; The predicted state coefficient is corrected according to the correction coefficient, and it is determined whether to generate an alarm instruction for the turbine blades according to the corrected predicted state coefficient.

2. The method for evaluating the condition of a steam turbine blade according to claim 1, wherein: The historical monitoring logs are divided according to the status type to obtain multiple monitoring log sets, including: Obtaining a historical state coefficient of a steam turbine blade in each historical monitoring log, comparing the historical state coefficient with a standard state coefficient interval when the steam turbine blade is in normal operation, and determining a state type of the steam turbine blade in each historical monitoring log according to the comparison result; If the historical state coefficient is within the standard state coefficient range, the state type of the turbine blade in the corresponding historical monitoring log is set to a normal state; If the historical state coefficient is not in the standard state coefficient interval, calculate the difference between the historical state coefficient and the historical state coefficient in the standard state coefficient interval; When the historical state coefficient difference is within the first preset difference interval, the state type of the steam turbine blade in the corresponding historical monitoring log is set as a suspected abnormal state; If the historical state coefficient difference is within the second preset difference interval, the state type of the steam turbine blade in the corresponding historical monitoring log is set to an abnormal state; Wherein, the first preset difference interval is smaller than the second preset difference interval; Several historical monitoring logs are divided according to their status types to obtain multiple monitoring log sets, which are respectively a monitoring log set of a normal state, a monitoring log set of an abnormal state, and a monitoring log set of a suspected abnormal state.

3. The method for evaluating the condition of a steam turbine blade according to claim 2, wherein: Analyze the first historical monitoring data to determine the focus factors and corresponding weight coefficients, including: Determine a normal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set in the normal state; Determine the abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set of abnormal state; Determine the suspected abnormal monitoring data interval corresponding to the first historical monitoring data according to the same first historical monitoring data of different historical monitoring logs in the monitoring log set of suspected abnormal state; Compare the normal monitoring data interval of each first historical monitoring data with the corresponding abnormal monitoring data interval and the suspected abnormal monitoring data interval, respectively, to determine the first data difference and the second data difference corresponding to the first historical monitoring data; If the first data difference of the first historical monitoring data is greater than the second data difference and the second data difference is greater than the preset data difference threshold, then the corresponding first historical monitoring data is set as a pending concern factor; Calculate the first pending data difference between the first historical feature data of each historical monitoring log in the monitoring log set in the normal state and the second historical feature data of each historical monitoring log in the monitoring log set in the suspected abnormal state, and the second pending data difference between the third historical feature data of each historical monitoring log in the monitoring log set in the abnormal state for each pending concern factor; Setting the accuracy of the pending concern factor according to a first number of historical monitoring logs whose first pending data difference is greater than a preset data difference threshold and a second number of historical monitoring data whose second pending data difference is greater than the first pending data difference; If the accuracy is greater than the preset accuracy threshold, the pending attention factor is set as the attention factor; A weight coefficient of the corresponding attention factor is generated according to a first attention difference value whose multiple first pending data difference values ​​of each attention factor are greater than a preset data difference threshold, a second attention difference value whose multiple second pending data difference values ​​are greater than the first pending data difference, and accuracy.

4. The method for evaluating the condition of a steam turbine blade according to claim 3, wherein: The calculation formula for the accuracy of the undetermined attention factor is: ； Where Z is the accuracy, is the first accurate conversion coefficient, is the second accurate conversion coefficient, d1 is the first number of historical monitoring logs whose first pending data difference is greater than the preset data difference threshold, m2 is the total number of historical monitoring logs in the monitoring log set of suspected abnormal state, d2 is the second number of historical monitoring data whose second pending data difference is greater than the first pending data difference, and m3 is the total number of historical monitoring logs in the monitoring log set of abnormal state; The calculation formula of the weight coefficient of the attention factor is: ； Among them, Q1 is the weight coefficient of the attention factor, q1 is the weight conversion coefficient, is the i1th first attention difference, a1 is the weight coefficient of the first attention difference, is the i2th second attention difference, and a2 is the weight coefficient of the second attention difference.

5. The method for evaluating the condition of a steam turbine blade according to claim 4, characterized in that: Determine the correlation factors and corresponding weight coefficients of the focus factors, including: The time length of the historical monitoring cycle of each historical monitoring log is used as a time reference line, multiple data collection nodes are set according to a preset time interval, multiple attention factors and second historical monitoring data in the corresponding historical monitoring log are obtained according to the data collection nodes, and mapped to the time reference line to obtain a data analysis graph of the corresponding historical monitoring log; The data analysis graph includes a data change curve of each concern factor in the corresponding historical monitoring log and a data change curve of each second historical monitoring data; Marking the mutation time node of the data change curve of each concerned factor in the data analysis graph, intercepting the first curve segment of the data change curve of each second historical monitoring data in the preceding adjacent time period of the marked mutation time node of each concerned factor, wherein each concerned factor has at least one marked mutation time node; Mapping the first curve segment of each second historical monitoring data with the marked mutation time node of each concerned factor to obtain a first curve segment-mutation time node mapping table of each second historical monitoring data for each concerned factor; Calculate the fluctuation degree of each second historical monitoring data in the first curve segment; If the fluctuation degree of all first curve segments in the first curve segment-mutation time node mapping table for the same focus factor of the same second historical monitoring data is greater than the preset fluctuation degree threshold, the second historical monitoring data is set as the correlation factor of the corresponding focus factor.

6. The method for evaluating the condition of a steam turbine blade according to claim 5, wherein: Determine the correlation factors and corresponding weight coefficients of the focus factors, including: Obtain the correlation factor of each concern factor, as well as the mapping relationship between the mutation degree at the annotated mutation time node of each concern factor and the fluctuation degree of the correlation factor in the monitoring log set of different state types, quantify the mutation degree and fluctuation degree with mapping relationship, and obtain the mutation quantization value and the corresponding fluctuation quantization value; According to the mutation quantization value of each concern factor and the corresponding correlation factor in different state types and the corresponding fluctuation quantization value, a first fluctuation quantization value-mutation quantization value mapping table, a second fluctuation quantization value-mutation quantization value mapping table and a third fluctuation quantization value-mutation quantization value mapping table are respectively constructed; Determine the first influence coefficient of each associated factor in a normal state on the corresponding concern factor according to the first fluctuation quantization value-mutation quantization value mapping table; Determine the second influence coefficient of each associated factor in the suspected abnormal state on the corresponding concern factor according to the second fluctuation quantization value-mutation quantization value mapping table; Determine the third influence coefficient of each associated factor under the abnormal state on the corresponding concern factor according to the third fluctuation quantization value-mutation quantization value mapping table; Determine the comprehensive influence coefficient of each correlation factor on the corresponding concern factor according to the first influence coefficient, the second influence coefficient and the third influence coefficient; Generate a weight coefficient of the corresponding correlation factor according to the number of attention factors corresponding to the same correlation factor and the comprehensive influence coefficient of the correlation factor on each corresponding attention factor; The calculation formula of the weight coefficient of the association factor is: ； Among them, Q2 is the weight coefficient of the correlation factor, q2 is the weight conversion coefficient of the correlation factor, n1 is the number of attention factors corresponding to the current correlation factor, n2 is the total number of attention factors, and Ys is the comprehensive influence coefficient of the correlation factor on the sth attention factor.

7. The method for evaluating the condition of a steam turbine blade according to claim 6, wherein: A state assessment model is constructed based on the focus factors and the corresponding weight coefficients, and an auxiliary assessment model is constructed based on the correlation factors of the focus factors and the corresponding weight coefficients, including: According to the attention factors and corresponding weight coefficients of historical monitoring logs in different status types as training input data, and the historical status coefficients of the corresponding historical monitoring logs as training output data, neural network training is performed to obtain a status assessment model for steam turbine blades; According to the correlation factor of each concern factor, the fluctuation degree of the correlation factor and the weight coefficient of the correlation factor as the training input data, and the mutation degree of the corresponding concern factor as the training output data, neural network training is carried out to obtain the auxiliary evaluation model of turbine blades.

8. The method for evaluating the condition of a steam turbine blade according to claim 7, wherein: The predicted state coefficient of the steam turbine blade at the current monitoring time node is generated according to the state assessment model, and the correction coefficient is obtained according to the auxiliary assessment model, including: Obtain the real-time attention factor of the current monitoring time node, input the real-time attention factor and the corresponding weight coefficient into the state assessment model to obtain the predicted state coefficient; Obtain the historical correlation factor and the fluctuation degree of each concern factor in the previous preset period, and input them into the auxiliary evaluation model in combination with the weight coefficient of the historical correlation factor to obtain the predicted mutation degree of the corresponding concern factor; Generate the predicted attention factor of the current monitoring time node according to the historical attention factor of the previous preset period and the predicted mutation degree, compare the predicted attention factor with the corresponding real-time attention factor, and obtain the difference value; Generate a degree of deviation according to a plurality of difference magnitudes and weight coefficients of corresponding attention factors; The correction factor of the predicted state coefficient is set according to the degree of deviation.

9. The method for evaluating the condition of a steam turbine blade according to claim 8, characterized in that: Whether to generate an alarm instruction for the steam turbine blades is determined according to the corrected predicted state coefficient, including: Obtain the historical state coefficient of the previous monitoring time node, and determine the change characteristics of the state coefficient according to the predicted state coefficient of the current monitoring time node and the historical state coefficient of the previous monitoring time node, wherein the change characteristics include the change trend, change rate and change value; Generate a compensation coefficient according to the change trend, change rate and change value of the state coefficient; Presetting a state coefficient threshold and a compensation coefficient threshold; When the corrected predicted state coefficient is less than the state coefficient threshold, an alarm instruction is sent; When the corrected predicted state coefficient is not less than the state coefficient threshold, and the compensation coefficient is less than the compensation coefficient threshold, a warning signal is sent; When the corrected predicted state coefficient is not less than the state coefficient threshold and the compensation coefficient is greater than the compensation coefficient threshold, no alarm instruction and no early warning instruction are sent.

Images