Turbine fault detection method and system based on big data

Abstract

The invention relates to the technical field of mechanical vibration measurement, in particular to a steam turbine fault detection method and system based on big data, and the method comprises the steps: obtaining a first real-time vibration value of a blade at a first moment in the operation process of a steam turbine; predicting a first predicted vibration value of the blade at the first moment through the initial vibration prediction model; determining a first possibility that the blade is abnormal in the vibration data segment; adjusting the initial vibration prediction model according to the first predicted vibration value, the first possibility and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; predicting a second predicted vibration value of the blade at the first moment in the running process of the steam turbine by utilizing the vibration prediction model; and detecting the fault of the steam turbine according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment. Therefore, the accuracy of the fault detection result of the steam turbine can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical vibration measurement, and in particular to a steam turbine fault detection method and system based on big data. Background Art

[0002] A steam turbine is a device that converts the thermal energy of steam into mechanical energy. The steam expands and rotates on the impeller to drive the rotating shaft and output power. Steam turbines are widely used in generators, propellers or other mechanical devices in power plants, ships and industrial equipment. The blades on the steam turbine are one of its important structures, and their normal operation is crucial to the safety, efficiency, stability and economy of the equipment. Therefore, steam turbine blade fault detection has become an important part of steam turbine fault detection.

[0003] In some scenarios, the method of detecting faults in steam turbine blades mainly relies on monitoring the vibration data of the blades during the operation of the steam turbine. First, the vibration data of the blades during operation is collected in real time through sensors or other monitoring equipment installed on the blades, and then a prediction model is established using historical data. Finally, the real-time monitored vibration data is compared with the prediction model. If the monitored vibration data deviates from the expected value of the prediction model, it is considered that there may be a fault. However, during the operation of the steam turbine, operating parameters such as steam pressure, temperature, and flow are often adjusted according to power demand. The prediction model often cannot adapt well to the vibration of the blades after these parameters are adjusted, resulting in inaccurate fault detection results for the steam turbine. Summary of the invention

[0004] In order to solve the technical problem of low accuracy of turbine fault detection results, the purpose of the present invention is to provide a turbine fault detection method and system based on big data. The technical solution adopted is as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for detecting turbine faults based on big data, including: obtaining a first real-time vibration value of a blade at a first moment during operation of the turbine; inputting the first real-time vibration value into an initial vibration prediction model, and predicting a first predicted vibration value of the blade at the first moment through the initial vibration prediction model; determining a first possibility that an abnormality occurs in the blade in the vibration data segment according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment, the first possibility characterizing the possibility that the abnormality occurs in the blade due to a change in an operating parameter of the turbine, the vibration data segment including a first moment and multiple moments after the first moment, and the historical data segment including multiple historical moments between the first moment; adjusting the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; predicting a second predicted vibration value of the blade at a first moment during operation of the turbine using the vibration prediction model; and detecting a turbine fault according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment.

[0006] Optionally, before inputting the first real-time vibration value into the initial vibration prediction model and predicting the first predicted vibration value of the blade at the first moment through the initial vibration prediction model, the method also includes: obtaining the second real-time vibration value of the turbine blade at each moment within a predetermined time period; removing abnormal values ​​in the second real-time vibration value at each moment within the predetermined time period; supplementing data at the position corresponding to the abnormal value to obtain the third real-time vibration value at the position corresponding to the abnormal value; and establishing an initial vibration prediction model for the blade based on the second real-time vibration value and the third real-time vibration value.

[0007] Optionally, establishing an initial vibration prediction model for the blade based on the second real-time vibration value and the third real-time vibration value includes: using the least squares method to perform curve fitting on the second real-time vibration value and the third real-time vibration value to obtain a vibration data fitting curve; obtaining extreme points in the vibration data fitting curve; determining a second possibility that the extreme point is an extreme point of the vibration curve during the turbine startup process based on the second real-time vibration value or the third real-time vibration value corresponding to the adjacent extreme points; determining the first target extreme point during the turbine non-startup process based on the second possibility; segmenting the vibration data fitting curve based on the time at which the second target extreme point, which is farthest from the time corresponding to the first target extreme point during the turbine non-startup process, is located; and establishing an initial vibration prediction model for the blade based on the second real-time vibration value and / or the third real-time vibration value after the moment corresponding to the second target extreme point.

[0008] Optionally, removing abnormal values ​​in the second real-time vibration value at each moment within a predetermined time period includes: determining an average vibration value of the second real-time vibration value within the predetermined time period; normalizing the absolute value of the first difference between the second real-time vibration value at each moment and the average vibration value to obtain the degree of abnormality of the second real-time vibration value at each moment; determining the second real-time vibration value whose degree of abnormality is greater than a first threshold as an abnormal value and removing it.

[0009] Optionally, according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value and the historical vibration value at each moment in the historical data segment, determining the first possibility that the blade has an abnormality in the vibration data segment includes: calculating the absolute value of the second difference between the first real-time vibration value and the first predicted vibration value, and the absolute value of the third difference between the first predicted vibration value and the blade vibration data of the initial vibration prediction model, the blade vibration data including all predicted vibration values ​​predicted by the initial vibration prediction model; calculating a first ratio of the maximum value of the absolute value of the second difference to the absolute value of the third difference, and determining the first ratio as the degree of deviation of the first real-time vibration value at the first moment relative to the first predicted vibration value; when the degree of deviation is greater than a second threshold value, determining the first real-time vibration value at the first moment The vibration value at the time is suspected vibration abnormality data; for the suspected vibration abnormality data, according to the absolute value of the first deviation value between the first real-time vibration value of the current blade of the steam turbine and the first predicted vibration value, and the absolute value of the second deviation value between the first real-time vibration value of other blades of the steam turbine and the first predicted vibration value, determine the third possibility that the current blade is abnormal at the first moment, and the third possibility represents the possibility that the abnormality of the blade is caused by a blade failure; when the third possibility is greater than or equal to the third threshold, according to the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at the corresponding moment in the historical data segment, determine the similarity of the vibration data between the vibration data segment and the historical data segment; based on the similarity, determine the first possibility that the blade is abnormal in the vibration data segment.

[0010] Optionally, before determining the similarity of vibration data between the vibration data segment and the historical data segment based on the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at the corresponding moment in the historical data segment, the method also includes: determining the position of the first real-time vibration value corresponding to the suspected vibration abnormality data in the vibration data segment based on the first real-time vibration value corresponding to the suspected vibration abnormality data, the maximum vibration value and the minimum vibration value in the vibration data segment; and selecting a historical data segment from the historical data with the position as the starting position and consistent with the duration of the vibration data segment.

[0011] Optionally, the initial vibration prediction model is adjusted according to the first predicted vibration value, the first possibility and the first real-time vibration value at each moment in the vibration data segment to obtain the vibration prediction model, including: calculating the first product between the first possibility and the first real-time vibration value at each moment in the vibration data segment, and the second product between the preset value and the fifth difference of the first possibility and the first predicted vibration value at each moment in the vibration data segment; determining the sum of the first product and the second product as the adjusted third predicted vibration value; and establishing the vibration prediction model based on the third predicted vibration value at each moment.

[0012] Optionally, detecting a fault in a steam turbine based on a first real-time vibration value and a second predicted vibration value of the blade at a first moment includes: calculating the absolute value of a sixth difference between the first real-time vibration value and the second predicted vibration value at the first moment; normalizing the absolute value of the sixth difference to obtain a degree of vibration abnormality of the blade; and determining that a fault exists in the blade of the steam turbine when the degree of vibration abnormality is greater than a fourth threshold value.

[0013] In a second aspect, an embodiment of the present application provides a turbine fault detection system based on big data, including: an acquisition module, used to acquire a first real-time vibration value of a blade at a first moment during the operation of the turbine; a prediction module, used to input the first real-time vibration value into an initial vibration prediction model, and predict the first predicted vibration value of the blade at the first moment through the initial vibration prediction model; a determination module, used to determine a first possibility that an abnormality occurs in the blade in the vibration data segment according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment, the first possibility characterizing the possibility that the abnormality of the blade is caused by a change in the operating parameters of the turbine, the vibration data segment including the first moment and multiple moments after the first moment, and the historical data segment including multiple historical moments between the first moment; an adjustment module, used to adjust the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; the prediction module is also used to predict a second predicted vibration value of the blade at the first moment during the operation of the turbine using the vibration prediction model; a detection module, used to detect a fault of the turbine according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment.

[0014] In the third aspect, an embodiment of the present application provides a steam turbine fault detection system based on big data, which comprises: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; and the processor is used to execute the program stored in the memory to implement the steps of the steam turbine fault detection method based on big data as mentioned in the first aspect.

[0015] The present invention has the following beneficial effects: firstly, a first real-time vibration value of a blade at a first moment during the operation of a steam turbine is obtained; then, the first real-time vibration value is input into an initial vibration prediction model, and a first predicted vibration value of the blade at the first moment is predicted by the initial vibration prediction model; and according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value and the historical vibration value at each moment in the historical data segment, a first possibility of an abnormality of the blade in the vibration data segment is determined, the first possibility characterizing the possibility that the abnormality of the blade is caused by a change in the operating parameters of the steam turbine, the vibration data segment includes a first moment and multiple moments after the first moment, and the historical data segment includes multiple historical moments between the first moment; secondly, the initial vibration prediction model is adjusted according to the first predicted vibration value, the first possibility and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; and then the vibration prediction model is used to predict a second predicted vibration value of the blade at the first moment during the operation of the steam turbine; finally, the fault of the steam turbine is detected according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment.

[0016] In this way, the embodiment of the present invention adjusts the initial vibration prediction model through the possibility that the abnormality of the turbine blades is caused by changes in the operating parameters of the turbine, which can effectively improve the adaptability of the vibration prediction model after the operating parameters of the turbine are adjusted, thereby improving the accuracy of the turbine fault detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a steam turbine fault detection method based on big data provided by one embodiment of the present invention;

[0019] Figure 2 A schematic structural diagram of a steam turbine blade provided by one embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the module composition of a steam turbine fault detection system based on big data provided by one embodiment of the present invention;

[0021] Figure 4 A schematic structural diagram of a steam turbine fault detection system based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a method and system for detecting a steam turbine fault based on big data proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0023] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0024] The specific scheme of a steam turbine fault detection method and system based on big data provided by the present invention is described in detail below with reference to the accompanying drawings.

[0025] Embodiment 1:

[0026] See also Figure 1 , which shows a flow chart of a steam turbine fault detection method based on big data provided by an embodiment of the present invention, including:

[0027] S101, obtaining a first real-time vibration value of a blade at a first moment during operation of a steam turbine.

[0028] Specifically, the blades of the steam turbine are Figure 2 As shown, Figure 2 A schematic diagram of the structure of a steam turbine blade provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, one end of the turbine blade is fixed and the other end is not fixed. In some scenarios, the vibration amplitude is the largest in the middle of the lower half of the fixed point of the turbine blade, which can best show the abnormal vibration of the blade. Therefore, a vibration sensor can be installed in the middle of the lower half of the fixed point of each turbine blade, and the monitoring frequency of the vibration sensor can be set to 1000HZ.

[0029] Furthermore, the steam turbine has multiple blades, and a vibration sensor is installed for each blade, and the vibration sensor can collect the real-time vibration value of each blade at each moment in the monitoring period. The first moment can be the current moment. The first real-time vibration value can be the real-time vibration amplitude of the blade.

[0030] S102: inputting a first real-time vibration value into an initial vibration prediction model, and predicting a first predicted vibration value of the blade at a first moment by using the initial vibration prediction model.

[0031] Specifically, the vibration of the turbine blades conforms to the periodic law, so the vibration of the turbine blades can be predicted based on the initial vibration prediction model. If the difference between the predicted vibration value at the first moment predicted by the initial vibration prediction model and the first real-time vibration value is too large, it means that the real-time vibration data of the blades may be abnormal. The abnormal reason for the real-time vibration data may be that the blades may be abnormal or the parameters of the turbine have changed. Therefore, it is necessary to distinguish the abnormalities in the real-time vibration data caused by the parameters of the turbine in order to identify the real fault of the blades.

[0032] Furthermore, the initial vibration prediction model may be pre-established. In an optional embodiment of the present invention, when establishing the initial vibration prediction model, firstly, the second real-time vibration value of the turbine blade at each moment within a predetermined time period is obtained; then, the abnormal values ​​in the second real-time vibration value at each moment within the predetermined time period are removed; then, data is supplemented at the position corresponding to the abnormal value to obtain the third real-time vibration value at the position corresponding to the abnormal value; finally, the initial vibration prediction model of the blade is established according to the second real-time vibration value and the third real-time vibration value.

[0033] Specifically, the predetermined time period can be determined according to actual conditions. In the embodiment of the present invention, the predetermined time period can be set to 5 minutes, and the interval between adjacent moments within 5 minutes can be 1 second. Therefore, the embodiment of the present invention continuously monitors the second real-time vibration value of the blades of the steam turbine within 5 minutes. Due to the influence of the external environment, there may be abnormal values ​​in the second real-time vibration value of the steam turbine at different moments within the predetermined time period. In order to avoid affecting the prediction accuracy of the initial vibration prediction model, the embodiment of the present invention eliminates the abnormal values ​​within the predetermined time period.

[0034] Furthermore, when eliminating abnormal values ​​of the second real-time vibration value within a predetermined time period, as an optional embodiment of the present invention, the average vibration value of the second real-time vibration value within the predetermined time period is first determined; then the absolute value of the first difference between the second real-time vibration value and the average vibration value at each moment is normalized to obtain the abnormality degree of the second real-time vibration value at each moment; finally, the second real-time vibration value with an abnormality degree greater than the first threshold is determined as an abnormal value and removed.

[0035] Specifically, the abnormality degree of the second real-time vibration value at each moment can be expressed by the following formula:

[0036]

[0037] In the above formula, e i Indicates the abnormality of the ith second real-time vibration value, z i represents the i-th second real-time vibration value, represents the average vibration value of the second real-time vibration value in a predetermined time period corresponding to the i-th second real-time vibration value. norm is a normalization function, which is used to Normalize so that e i The value is in the range of 0 to 1.

[0038] Furthermore, the first threshold value can be determined according to actual conditions, and in the embodiment of the present invention, the value is 0.7. i >0.7, the second real-time vibration value at that moment is considered to be an abnormal value and needs to be eliminated.

[0039] Furthermore, when supplementing data at the position corresponding to the abnormal value, as an optional embodiment of the present invention, at the position where the abnormal value is removed, the average of the two second real-time vibration values ​​before and after the position can be used as the third real-time vibration value of the position. The second real-time vibration value and the third real-time vibration value can both be the vibration amplitude of the blade.

[0040] Furthermore, there are multiple blades in the steam turbine. In the embodiment of the present invention, an initial vibration prediction model corresponding to each blade is established. When the initial vibration prediction model of the blade is established for each blade, as an optional embodiment of the present invention, the second real-time vibration value and the third real-time vibration value are firstly curve-fitted using the least squares method to obtain a vibration data fitting curve; then the extreme value points in the vibration data fitting curve are obtained; then, according to the second real-time vibration values ​​or the third real-time vibration values ​​corresponding to the adjacent extreme value points, the second possibility that the extreme value point is the extreme value point of the vibration curve during the startup process of the steam turbine is determined; secondly, according to the second possibility, the first target extreme value point during the non-starting process of the steam turbine is determined; and the vibration data fitting curve is segmented using the time at which the second target extreme value point, which is the farthest in time from the time corresponding to the first target extreme value point during the non-starting process of the steam turbine, is located; finally, the initial vibration prediction model of the blade is established based on the second real-time vibration value and / or the third real-time vibration value after the moment corresponding to the second target extreme value point.

[0041] Specifically, the real-time vibration value processed by the above embodiment of the present invention is subjected to curve fitting using the least square method. Since the real-time vibration value after the above processing may include the vibration value during the turbine startup process, and the blades have not yet reached the stable vibration stage during the turbine startup process, it is necessary to remove the vibration value during the turbine startup process when establishing the initial vibration prediction model of the blades. When removing the vibration value during the turbine startup process, the extreme value point in the vibration data fitting curve can be used to determine whether the turbine is in the startup process. The extreme value point can be a maximum value point and a minimum value point. The extreme value point is selected to determine whether the turbine is in the startup process. This is because when the blades of the turbine are in a stable state, the extreme value point in the vibration data fitting curve should also be relatively stable. In the specific implementation process, the possibility that each extreme value point is an extreme value point in the vibration curve during the turbine startup process can be calculated based on the difference between each extreme value point and the two extreme value points before and after. It is worth noting that, according to the type of extreme point, if the current extreme point is a maximum point, the difference between the two adjacent maximum points before and after is used to calculate the possibility that the maximum point is an extreme point in the vibration curve during the turbine startup process. If the current extreme point is a minimum point, the difference between the two adjacent minimum points before and after is used to calculate the possibility that the minimum point is an extreme point in the vibration curve during the turbine startup process.

[0042] Furthermore, the second possibility can be expressed as follows:

[0043] q a =norm(|z a -z a-1 |+|z a+1 -z a |)

[0044] Upper middle, q a Indicates the second possibility that the ath extreme point is the extreme point of the vibration curve during the turbine startup process. a Indicates the real-time vibration value of the ath extreme point. a-1 Represents the real-time vibration value of the extreme point at the previous moment adjacent to the ath extreme point. a+1 represents the real-time vibration value of the extreme point at the next moment adjacent to the ath extreme point. norm represents the normalization function, which is used to |z a -z a-1 |+|z a+1 -z a |Perform normalization.

[0045] Furthermore, after determining the second possibility that the extreme point at each moment in the vibration data fitting curve is the extreme point of the vibration curve during the turbine startup process, the extreme point whose second possibility is greater than or equal to the fourth threshold can be selected as the extreme point during the turbine startup process, and the extreme point whose second possibility is less than the fourth threshold can be selected as the extreme point during the turbine non-startup process. Among them, the fourth threshold can be determined according to actual conditions, and in the embodiment of the present invention, the value is 0.6. When the second possibility q of a certain extreme point a When ≥0.6, the extreme point is considered to be the extreme point in the turbine startup process, otherwise, it is the extreme point in the turbine non-startup process. The extreme point with the farthest (earliest) time corresponding to the extreme point in the turbine non-startup process is taken to separate the vibration data fitting curve. The vibration data fitting curve before the time corresponding to the extreme point is the blade vibration curve in the turbine startup process, and the vibration data fitting curve after the time corresponding to the extreme point is the initial vibration prediction model.

[0046] S103, determining a first possibility that an abnormality occurs in the blade in the vibration data segment according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment.

[0047] Among them, the first possibility represents the possibility that the abnormality of the blade is caused by the change of the operating parameters of the turbine, the vibration data segment includes the first moment and multiple moments after the first moment, and the historical data segment includes multiple historical moments between the first moment.

[0048] Specifically, after the initial vibration prediction model of the blade is established through the above-mentioned embodiment of the present invention, the initial vibration prediction model is used to predict the first predicted vibration value of the blade at each moment. Among them, the vibration data segment is a data segment used to analyze the vibration law of the abnormal first real-time vibration value. The length of the vibration data segment can be continuously monitored for 5 seconds from the first moment, that is, 5 moments after the first moment. The real-time vibration data of the blade within the 5 seconds is used as the vibration data segment corresponding to the first real-time vibration data at the first moment. The historical time period refers to the data segment in the historical data that is consistent with the operating parameters of the steam turbine corresponding to the first moment. Among them, the operating parameters of the steam turbine include but are not limited to steam pressure, steam flow, turbine speed, exhaust pressure, etc. The length of the historical data segment can be consistent with the length of the vibration data segment.

[0049] Further, when determining the first possibility that the blade has an abnormality in the vibration data segment, as an optional embodiment of the present invention, firstly, the absolute value of the second difference between the first real-time vibration value and the first predicted vibration value, and the absolute value of the third difference between the first predicted vibration value and the blade vibration data of the initial vibration prediction model are calculated, and the blade vibration data includes all predicted vibration values ​​predicted by the initial vibration prediction model; and the first ratio of the maximum value of the absolute value of the second difference to the absolute value of the third difference is calculated, and the first ratio is determined to be the degree of deviation of the first real-time vibration value at the first moment relative to the first predicted vibration value; when the degree of deviation is greater than the second threshold value, it is determined that the first real-time vibration value at the first moment is suspected vibration abnormality data; and secondly, for For suspected vibration abnormality data, a third possibility that the current blade has an abnormality at the first moment is determined based on the absolute value of the first deviation value between the first real-time vibration value of the current blade of the turbine and the first predicted vibration value, and the absolute value of the second deviation value between the first real-time vibration value of other blades of the turbine and the first predicted vibration value. The third possibility represents the possibility that the abnormality of the blade is caused by a blade failure. When the third possibility is greater than or equal to the third threshold, the similarity of vibration data between the vibration data segment and the historical data segment is determined based on the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at the corresponding moment in the historical data segment. Finally, based on the similarity, the first possibility that the blade has an abnormality in the vibration data segment is determined.

[0050] Specifically, the degree of deviation of the first real-time vibration value at the first moment relative to the first predicted vibration value can be expressed by the following formula:

[0051]

[0052] In the above formula, d represents the deviation degree of the first real-time vibration value of the currently monitored blade at the first moment relative to the first predicted vibration value. z represents the first real-time vibration value of the currently monitored blade at the first moment. pre represents the first predicted vibration value of the blade currently being monitored at the first moment. i represents the blade vibration data of the initial vibration prediction model. max(|z pre -z i |) represents the maximum deviation value relative to the first moment within the prediction range of the initial vibration prediction model. |z pre -z| represents the deviation value of the first real-time vibration value of the blade currently monitored at the first moment relative to the first predicted vibration value.

[0053] Further, the second threshold value can be taken according to the actual situation, and the value is 0.7 in the embodiment of the present invention. When the deviation degree d of the first real-time vibration value at a certain moment is greater than 0.7, it is considered that the first real-time vibration value monitored at this moment is suspected vibration abnormal data. Among them, the reason for the appearance of the suspected vibration abnormal data may be caused by turbine failures such as blade damage and mechanical failure, or it may be caused by the change of blade vibration caused by the adjustment of the working parameters of the turbine, thereby causing the vibration value of the blade to be abnormal. When the abnormality is caused by the adjustment of the working parameters of the turbine, it means that the initial vibration prediction model of the current blade does not conform to the blade vibration law under the current turbine working parameters, that is, the initial vibration prediction model of the blade needs to be adjusted. Therefore, it is necessary to further distinguish whether the abnormal cause of the suspected vibration abnormal data is caused by the abnormal turbine failure or the adjustment of the working parameters of the turbine. Among them, when a single blade has an abnormality such as a crack, the vibration abnormality will occur on the blade with the abnormality, and then it may be transmitted to other blades as the blade vibration affects the bearing and the rotor. Therefore, when the abnormality of a single blade just occurs, only a small number of blades will show vibration abnormality. Therefore, the possibility that the suspected abnormal vibration data is a single blade abnormality can be calculated based on the similarity of the degree of deviation of the vibration values ​​of all blades relative to the predicted values ​​at the time corresponding to the suspected abnormal vibration data.

[0054] Furthermore, the third possibility can be expressed as follows:

[0055]

[0056] In the above formula, p d Indicates the third possibility that the current blade is abnormal at the first moment. p Indicates the number of other blades on the turbine except the current blade. u indicates the u-th blade among the other blades of the turbine. Δz u represents the second deviation value between the real-time vibration value of the u-th blade in the other blades of the steam turbine at the first moment and the predicted vibration value predicted by its initial vibration prediction model (a real-time vibration value greater than the predicted vibration value is positive, and a real-time vibration value less than the predicted vibration value is negative). Δz represents the first deviation value between the first real-time vibration value of the current blade and the first predicted vibration value. norm represents the normalization function, which is used to Normalize.

[0057] Furthermore, p d The larger the value is, the more likely it is that the abnormal cause of the suspected abnormal vibration data is caused by an abnormality on a single blade. dThe smaller the value is, the more likely the abnormal cause of the suspected abnormal vibration data is caused by the change of turbine operating parameters. The third threshold value can be determined according to actual conditions, and the value is 0.4 in the embodiment of the present invention. d When p ≥0.4, it is considered that the abnormal cause of the suspected abnormal vibration data is caused by a fault on a single blade, indicating that there is an abnormality on the turbine blade, and there is no need to adjust the prediction model, but the turbine fault needs to be repaired. d When <0.4, it is necessary to further analyze whether the abnormal cause of the suspected abnormal vibration data is caused by changes in turbine operating parameters or other abnormal conditions such as overheating of the turbine.

[0058] Furthermore, when determining the similarity of vibration data between the vibration data segment and the historical data segment, the following formula may be used for calculation:

[0059]

[0060] In the above formula, s t Indicates the similarity between the tth historical data segment and the vibration data segment where the current suspected vibration abnormal data is located. i Indicates the number of data points contained in the vibration data segment and the historical data segment. y,i Indicates the first real-time vibration value of the ith data point in the vibration data segment where the current suspected vibration abnormality data is located. t,i Represents the historical vibration value of the i-th data point in the t-th historical data segment. The higher the similarity between the historical data segment and the vibration data segment where the current suspected abnormal vibration data is located, the higher the possibility that the suspected abnormal vibration data is caused by the change of the working parameters of the steam turbine.

[0061] Furthermore, when determining the historical data segment, as an optional embodiment of the present invention, the position of the first real-time vibration value corresponding to the suspected vibration abnormality data in the vibration data segment is first determined based on the first real-time vibration value corresponding to the suspected vibration abnormality data, the maximum vibration value and the minimum vibration value in the vibration data segment; and then a historical data segment with the position as the starting position and consistent with the duration of the vibration data segment is selected from the historical data.

[0062] Specifically, the position of the first real-time vibration value corresponding to the suspected abnormal vibration data in the vibration data segment can be expressed by the following formula:

[0063]

[0064] In the above formula, w represents the position of the first real-time vibration value corresponding to the suspected vibration abnormality data in the vibration data segment. z represents the first real-time vibration value corresponding to the suspected vibration abnormality data. minIndicates that the first real-time vibration value corresponding to the suspected vibration abnormality data is the minimum vibration value in the vibration data segment. max Indicates that the first real-time vibration value corresponding to the suspected vibration abnormality data is the maximum vibration value in the vibration data segment.

[0065] In the historical data whose operating parameters of the steam turbine are consistent with the operating parameters of the steam turbine at the current moment, the length of the vibration data segment is used as the length of the historical data segment. For example, if the length of the vibration data segment is 5 seconds, the length of the historical data segment is also 5 seconds. In the historical data, the length of the vibration data segment is used as the length of the historical data segment, and the starting position of the data segment is the position of the first real-time vibration value corresponding to the suspected vibration abnormal data calculated by the above embodiment of the present invention in the vibration data segment, and the historical data segment is selected.

[0066] Further, when determining the first possibility that the blade is abnormal in the vibration data segment based on similarity, as an optional embodiment of the present invention, the following formula can be used for calculation:

[0067]

[0068] In the above formula, P represents the first possibility that the current suspected abnormal vibration data is caused by the adjustment of turbine operating parameters. t Indicates the number of historical data segments. t represents the similarity between the tth historical data segment and the vibration data segment where the current suspected vibration abnormal data is located. norm represents the normalization function, which is used to Normalize it so that its value is in the range of 0 to 1. The larger P is, the more likely the suspected abnormal vibration data is caused by changes in the turbine operating parameters, that is, the more likely the current initial vibration prediction model is not suitable for the current turbine operating state.

[0069] S104, adjusting the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model.

[0070] Specifically, after the present invention determines P through the above embodiment, the larger P is, the more likely the suspected abnormal vibration data is caused by changes in the turbine operating parameters, that is, the more likely the current initial vibration prediction model is not suitable for the current turbine operating state. Therefore, P can be used as the weight of the suspected abnormal vibration data in the initial vibration prediction model to adjust the initial vibration prediction model.

[0071] Furthermore, when adjusting the initial vibration prediction model, as an optional embodiment of the present invention, firstly, a first product between the first possibility and the first real-time vibration value at each moment in the vibration data segment, and a second product between the preset value and the fifth difference between the first possibility and the first predicted vibration value at each moment in the vibration data segment are calculated; then, the sum of the first product and the second product is determined to be the adjusted third predicted vibration value; finally, a vibration prediction model is established based on the third predicted vibration value at each moment.

[0072] Specifically, the preset value can be determined according to actual conditions, and in the embodiment of the present invention, the value is 1. As an optional embodiment of the present invention, when determining the adjusted third predicted vibration value, the following formula can be used:

[0073] z v =P×z y,v +(1-P)×z pre,v

[0074] In the above formula, z v represents the third predicted vibration value after the adjustment of the first predicted vibration value at the vth moment. P represents the first possibility that the current suspected vibration abnormality data is caused by the adjustment of the turbine operating parameters. y,v Indicates the first real-time vibration value at the vth moment in the vibration data segment corresponding to the suspected vibration abnormality data. pre,v Represents the first predicted vibration value corresponding to the first real-time vibration value at the vth moment.

[0075] After obtaining the adjusted third predicted vibration value, a new vibration prediction model is generated. Specifically, a coordinate system is established for the adjusted third predicted vibration value in the vibration data segment according to "time-amplitude" to obtain its scatter plot. The third predicted vibration value in the vibration data segment is fitted using the least squares method, and the obtained fitting curve is the adjusted vibration prediction model.

[0076] S105, using a vibration prediction model to predict a second predicted vibration value of the blade at a first moment during operation of the steam turbine.

[0077] Specifically, after the adjusted vibration prediction model is obtained, the adjusted vibration prediction model is used to predict a second predicted vibration value of the blade at a first moment during the operation of the steam turbine.

[0078] S106, detecting a fault of the steam turbine according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment.

[0079] Specifically, the degree of abnormality of the vibration of the blades of the steam turbine can be calculated according to the degree of deviation between the first real-time vibration value at the first moment and the second predicted vibration value. When detecting a fault in the steam turbine, as an optional embodiment of the present invention, the absolute value of the sixth difference between the first real-time vibration value at the first moment and the second predicted vibration value is first calculated; then the absolute value of the sixth difference is normalized to obtain the degree of abnormal vibration of the blades; finally, when the degree of abnormal vibration is greater than the third threshold, it is determined that the blades of the steam turbine are faulty.

[0080] Furthermore, the degree of vibration abnormality can be expressed by the following formula:

[0081] E=norm(|zz pre ′ |)

[0082] In the above formula, E represents the abnormal vibration degree of the current blade. z represents the first real-time vibration value of the blade currently monitored. pre ′ represents the second predicted vibration value of the blade vibration at this moment by the adjusted vibration prediction model. norm() represents the normalization function, which is used to adjust |zz pre ′ |Normalize the values ​​so that they are in the range of 0 to 1.

[0083] Furthermore, the fourth threshold value can be set according to actual conditions, and in the embodiment of the present invention, the value is 0.3. When E≤0.3, it is considered that the blade vibration is within the normal range. When E>0.3, it is considered that the blade is faulty and the entire turbine needs to be repaired and adjusted.

[0084] The embodiment of the present invention adjusts the initial vibration prediction model through the possibility that the abnormality of the turbine blades is caused by the change of the operating parameters of the turbine, which can effectively improve the adaptability of the vibration prediction model after the operating parameters of the turbine are adjusted, thereby improving the accuracy of the turbine fault detection results.

[0085] Embodiment 2:

[0086] Corresponding to the steam turbine fault detection method based on big data provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a steam turbine fault detection system based on big data. Figure 3 A schematic diagram of the module composition of a steam turbine fault detection system based on big data provided in an embodiment of the present application, wherein the steam turbine fault detection system based on big data is used to execute Figure 1 The steam turbine fault detection method based on big data described in Figure 3As shown, the turbine fault detection system 300 based on big data includes: an acquisition module 301, which is used to acquire a first real-time vibration value of a blade at a first moment during the operation of the turbine; a prediction module 302, which is used to input the first real-time vibration value into an initial vibration prediction model, and predict the first predicted vibration value of the blade at the first moment through the initial vibration prediction model; a determination module 303, which is used to determine a first possibility that an abnormality occurs in the blade in the vibration data segment according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment, wherein the first possibility indicates that the abnormality of the blade is caused by the operation of the turbine. The possibility caused by the change of parameters, the vibration data segment includes the first moment and multiple moments after the first moment, and the historical data segment includes multiple historical moments between the first moment; the adjustment module 304 is used to adjust the initial vibration prediction model according to the first predicted vibration value, the first possibility and the first real-time vibration value of each moment in the vibration data segment to obtain the vibration prediction model; the prediction module 302 is also used to use the vibration prediction model to predict the second predicted vibration value of the blade at the first moment during the operation of the turbine; the detection module 305 is used to detect the fault of the turbine according to the first real-time vibration value and the second predicted vibration value of the blade at the first moment.

[0087] The embodiment of the present invention adjusts the initial vibration prediction model through the possibility that the abnormality of the turbine blades is caused by the change of the operating parameters of the turbine, which can effectively improve the adaptability of the vibration prediction model after the operating parameters of the turbine are adjusted, thereby improving the accuracy of the turbine fault detection results.

[0088] Optionally, the acquisition module 301 is also used to obtain the second real-time vibration value of the turbine blade at each moment within a predetermined time period; remove abnormal values ​​in the second real-time vibration value at each moment within the predetermined time period; supplement data at the position corresponding to the abnormal value to obtain the third real-time vibration value at the position corresponding to the abnormal value; and establish an initial vibration prediction model for the blade based on the second real-time vibration value and the third real-time vibration value.

[0089] Optionally, the prediction module 302 is also used to perform curve fitting on the second real-time vibration value and the third real-time vibration value using the least squares method to obtain a vibration data fitting curve; obtain extreme points in the vibration data fitting curve; determine a second possibility that the extreme point is an extreme point of the vibration curve during the turbine startup process based on the second real-time vibration value or the third real-time vibration value corresponding to the adjacent extreme points; determine the first target extreme point during the turbine non-startup process based on the second possibility; segment the vibration data fitting curve using the time at which the second target extreme point corresponding to the first target extreme point during the turbine non-startup process is located; and establish an initial vibration prediction model for the blade based on the second real-time vibration value and / or the third real-time vibration value after the moment corresponding to the second target extreme point.

[0090] Optionally, the prediction module 302 is also used to determine the average vibration value of the second real-time vibration value within a predetermined time period; normalize the absolute value of the first difference between the second real-time vibration value and the average vibration value at each moment to obtain the abnormality degree of the second real-time vibration value at each moment; determine the second real-time vibration value with an abnormality degree greater than a first threshold as an abnormal value and remove it.

[0091] Optionally, the determination module 303 is further used to calculate the absolute value of the second difference between the first real-time vibration value and the first predicted vibration value, and the absolute value of the third difference between the first predicted vibration value and the blade vibration data of the initial vibration prediction model, the blade vibration data including all predicted vibration values ​​predicted by the initial vibration prediction model; calculate a first ratio of the maximum value of the absolute value of the second difference to the absolute value of the third difference, and determine that the first ratio is the degree of deviation of the first real-time vibration value at the first moment relative to the first predicted vibration value; when the degree of deviation is greater than a second threshold value, determine that the first real-time vibration value at the first moment is suspected vibration abnormality data; for the suspected vibration abnormality data, according to the turbine The absolute value of the first deviation value between the first real-time vibration value of the current blade and the first predicted vibration value, and the absolute value of the second deviation value between the first real-time vibration value of other blades of the steam turbine and the first predicted vibration value are used to determine the third possibility that the current blade has an abnormality at the first moment, and the third possibility represents the possibility that the abnormality of the blade is caused by a blade failure; when the third possibility is greater than or equal to the third threshold, the similarity of vibration data between the vibration data segment and the historical data segment is determined according to the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at the corresponding moment in the historical data segment; based on the similarity, the first possibility that the blade has an abnormality in the vibration data segment is determined.

[0092] Optionally, the determination module 303 is also used to determine the position of the first real-time vibration value corresponding to the suspected vibration abnormality data in the vibration data segment based on the first real-time vibration value corresponding to the suspected vibration abnormality data, the maximum vibration value and the minimum vibration value in the vibration data segment; and select a historical data segment from the historical data that takes the position as the starting position and is consistent with the duration of the vibration data segment.

[0093] Optionally, the adjustment module 304 is also used to calculate a first product between the first possibility and the first real-time vibration value at each moment in the vibration data segment, and a second product between the preset value and the fifth difference between the first possibility and the first predicted vibration value at each moment in the vibration data segment; determine the sum of the first product and the second product as the adjusted third predicted vibration value; and establish a vibration prediction model based on the third predicted vibration value at each moment.

[0094] Optionally, the detection module 305 is also used to calculate the absolute value of the sixth difference between the first real-time vibration value at the first moment and the second predicted vibration value; normalize the absolute value of the sixth difference to obtain the degree of vibration abnormality of the blade; when the degree of vibration abnormality is greater than the fourth threshold, determine that there is a fault in the turbine blade.

[0095] The big data-based turbine fault detection system provided in the embodiment of the present application can implement the various processes in the embodiments corresponding to the above-mentioned big data-based turbine fault detection method, and has the same or similar beneficial effects. To avoid repetition, they will not be described here.

[0096] Embodiment three:

[0097] Corresponding to the steam turbine fault detection method based on big data provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a steam turbine fault detection system based on big data, and the steam turbine fault detection system based on big data is used to execute the above steam turbine fault detection method based on big data. Figure 4 A schematic diagram of a steam turbine fault detection system based on big data is provided to implement each embodiment of the present application. Figure 4 The turbine fault detection system based on big data may have relatively large differences due to different configurations or performances, and may include one or more processors 401 and memory 402, wherein the memory 402 is used to store computer programs that can be run on the processor 401, and the processor 401 is used to execute the programs stored in the memory 402 to achieve the above Figure 1 The various steps in the method embodiment. Among them, the memory 402 can be a temporary storage or a permanent storage. The application stored in the memory 402 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the steam turbine fault detection system based on big data.

[0098] Furthermore, the processor 401 can be configured to communicate with the memory 402, and execute a series of computer executable instructions in the memory 402 on the steam turbine fault detection system based on big data. The steam turbine fault detection system based on big data can also include one or more power supplies 403, one or more wired or wireless network interfaces 404, one or more input and output interfaces 405, and one or more keyboards 406.

[0099] Specifically in this embodiment, the turbine fault detection system based on big data includes a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store computer programs; the processor is used to execute the programs stored in the memory to achieve the above Figure 1 The various steps in the method embodiment are similar to those in the method embodiment, and have the beneficial effects of the above method embodiments. To avoid repetition, the embodiments of the present application will not be described again here.

[0100] It should be noted that the turbine fault detection system based on big data provided in the embodiment of the present application and the turbine fault detection method based on big data provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned turbine fault detection method based on big data, and has the same or similar beneficial effects, and the repeated parts will not be repeated.

[0101] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0102] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A steam turbine fault detection method based on big data, characterized in that: The steam turbine fault detection method based on big data includes: Obtaining a first real-time vibration value of a blade at a first moment during operation of a steam turbine; Inputting the first real-time vibration value into an initial vibration prediction model, and predicting a first predicted vibration value of the blade at the first moment by using the initial vibration prediction model; Determine, according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment, a first possibility that the blade has an abnormality in the vibration data segment, the first possibility representing the possibility that the abnormality of the blade is caused by a change in an operating parameter of the steam turbine, the vibration data segment including the first moment and multiple moments after the first moment, and the historical data segment including multiple historical moments between the first moment; Adjust the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; Using the vibration prediction model to predict a second predicted vibration value of the blade of the steam turbine at a first moment during operation; A fault of the steam turbine is detected based on a first real-time vibration value of the blade at a first moment and the second predicted vibration value.

2. The method for detecting steam turbine faults based on big data according to claim 1, characterized in that: Before inputting the first real-time vibration value into the initial vibration prediction model and predicting the first predicted vibration value of the blade at the first moment by using the initial vibration prediction model, the method further includes: Acquire a second real-time vibration value of the blade of the steam turbine at each moment within a predetermined time period; removing abnormal values ​​from the second real-time vibration value at each moment within the predetermined time period; Performing data supplementation at the position corresponding to the abnormal value to obtain a third real-time vibration value at the position corresponding to the abnormal value; An initial vibration prediction model of the blade is established according to the second real-time vibration value and the third real-time vibration value.

3. The method for detecting steam turbine faults based on big data according to claim 2, characterized in that: The step of establishing the initial vibration prediction model of the blade according to the second real-time vibration value and the third real-time vibration value comprises: Performing curve fitting on the second real-time vibration value and the third real-time vibration value by using a least square method to obtain a vibration data fitting curve; Obtaining extreme value points in the vibration data fitting curve; Determining a second possibility that the extreme point is an extreme point of the vibration curve during the startup of the steam turbine according to the second real-time vibration value or the third real-time vibration value corresponding to the adjacent extreme point; Determining a first target extreme value point in the non-startup process of the steam turbine according to the second possibility; The vibration data fitting curve is segmented using the time of the second target extreme point which is farthest from the time corresponding to the first target extreme point during the non-starting process of the steam turbine; An initial vibration prediction model of the blade is established based on the second real-time vibration value and / or the third real-time vibration value after the moment corresponding to the second target extreme value point.

4. The method for detecting steam turbine faults based on big data according to claim 2, characterized in that: The removing of abnormal values ​​in the second real-time vibration value at each moment within the predetermined time period comprises: determining an average vibration value of the second real-time vibration value within the predetermined time period; Normalizing the absolute value of the first difference between the second real-time vibration value at each moment and the average vibration value to obtain the abnormality degree of the second real-time vibration value at each moment; A second real-time vibration value whose abnormality level is greater than a first threshold is determined as the abnormal value and removed.

5. The method for detecting steam turbine faults based on big data according to any one of claims 1 to 4, characterized in that: Determining the first possibility that the blade has an abnormality in the vibration data segment according to the first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and the historical vibration value at each moment in the historical data segment includes: calculating an absolute value of a second difference between the first real-time vibration value and the first predicted vibration value, and an absolute value of a third difference between the first predicted vibration value and blade vibration data of the initial vibration prediction model, wherein the blade vibration data includes all predicted vibration values ​​predicted by the initial vibration prediction model; Calculating a first ratio of a maximum value of an absolute value of the second difference to an absolute value of the third difference, and determining the first ratio as a degree of deviation of a first real-time vibration value at the first moment relative to the first predicted vibration value; When the degree of deviation is greater than a second threshold, determining that the first real-time vibration value at the first moment is suspected vibration abnormality data; For the suspected vibration abnormality data, according to the absolute value of a first deviation value between a first real-time vibration value of a current blade of the steam turbine and the first predicted vibration value, and the absolute value of a second deviation value between a first real-time vibration value of other blades of the steam turbine and the first predicted vibration value, determine a third possibility that the current blade is abnormal at the first moment, wherein the third possibility represents the possibility that the abnormality of the blade is caused by a blade failure; In a case where the third possibility is greater than or equal to a third threshold, determining the similarity of vibration data between the vibration data segment and the historical data segment according to the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at a corresponding moment in the historical data segment; A first possibility that an abnormality occurs in the blade in the vibration data segment is determined based on the similarity.

6. The method for detecting steam turbine faults based on big data according to claim 5, characterized in that: Before determining the similarity of vibration data between the vibration data segment and the historical data segment based on the first real-time vibration value at each moment in the vibration data segment and the historical vibration value at a corresponding moment in the historical data segment, the method further includes: Determine the position of the first real-time vibration value corresponding to the suspected abnormal vibration data in the vibration data segment according to the first real-time vibration value corresponding to the suspected abnormal vibration data, the maximum vibration value and the minimum vibration value in the vibration data segment; A historical data segment with the position as the starting position and the same duration as the vibration data segment is selected from the historical data.

7. The method for detecting steam turbine faults based on big data according to claim 1, characterized in that: The step of adjusting the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain the vibration prediction model comprises: Calculating a first product between the first possibility and a first real-time vibration value at each moment in the vibration data segment, and a second product between a fifth difference between a preset value and the first possibility and a first predicted vibration value at each moment in the vibration data segment; determining a sum of the first product and the second product as an adjusted third predicted vibration value; The vibration prediction model is established based on the third predicted vibration value at each time.

8. The method for detecting steam turbine faults based on big data according to claim 1, characterized in that: The detecting of the fault of the steam turbine according to the first real-time vibration value of the blade at the first moment and the second predicted vibration value comprises: Calculate an absolute value of a sixth difference between the first real-time vibration value at the first moment and the second predicted vibration value; Normalizing the absolute value of the sixth difference to obtain the degree of abnormal vibration of the blade; In a case where the vibration abnormality level is greater than a fourth threshold value, it is determined that a blade of the steam turbine is faulty.

9. A steam turbine fault detection system based on big data, characterized in that: include: An acquisition module, used for acquiring a first real-time vibration value of a blade at a first moment during operation of the steam turbine; a prediction module, configured to input the first real-time vibration value into an initial vibration prediction model, and predict a first predicted vibration value of the blade at the first moment by using the initial vibration prediction model; a determination module, for determining a first possibility that the blade has an abnormality in the vibration data segment according to a first real-time vibration value at each moment in the vibration data segment, the first predicted vibration value, and a historical vibration value at each moment in the historical data segment, wherein the first possibility represents the possibility that the abnormality of the blade is caused by a change in an operating parameter of the steam turbine, the vibration data segment including the first moment and multiple moments after the first moment, and the historical data segment including multiple historical moments between the first moment; an adjustment module, configured to adjust the initial vibration prediction model according to the first predicted vibration value, the first possibility, and the first real-time vibration value at each moment in the vibration data segment to obtain a vibration prediction model; The prediction module is further used to predict a second predicted vibration value of the blade of the steam turbine at a first moment during operation by using the vibration prediction model; A detection module is used to detect a fault of the steam turbine according to a first real-time vibration value of the blade at a first moment and the second predicted vibration value.

10. A steam turbine fault detection system based on big data, characterized in that: The big data-based steam turbine fault detection system comprises: a processor and a memory; wherein the memory is used to store a computer program that can be run on the processor; and the processor is used to execute the program stored in the memory to implement the steps of the big data-based steam turbine fault detection method as described in any one of claims 1-8.

Images