Turbine troubleshooting method based on big data

Abstract

The invention provides a big data-based steam turbine fault elimination method, which comprises the following steps of: acquiring historical fault information of a steam turbine, and pre-classifying fault types based on industrial control information in the historical fault information; when the pre-classification result is an industrial control fault, outputting an industrial control fault correction strategy through a multi-stage analysis and judgment method; when the pre-classification result is a non-industrial control fault, calculating a priority sequence of typical faults in the big data statistical case based on a multi-criterion decision evaluation algorithm; and establishing a priority sequence-based troubleshooting strategy through a mapping method. The turbine fault is divided by taking the fault caused by industrial control as a starting point, the fault type is pre-analyzed from an equipment layer and human factors, and the priority ranking is performed in combination with typical fault cases of the historical turbine, so that the corresponding troubleshooting strategies are executed in sequence, the fault loss is reduced, and the fault detection efficiency is improved. And an execution sequence reference is provided for troubleshooting operation under the condition of multi-fault concurrency.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a steam turbine fault troubleshooting method based on big data. Background Art

[0002] Driven by the continuous rise in global energy demand and increasingly stringent environmental protection requirements, the power generation industry is in a critical period of profound change and transformation. With the iterative upgrade of power generation technology, modern power plants have started the process of upgrading and transforming in the direction of large capacity and high parameters in order to meet the huge demand for electricity in society and enhance their competitiveness in the energy market.

[0003] Large-capacity units, with their advantages of large-scale production, can significantly increase power generation per unit time and meet the growing social electricity demand. High-parameter operation is to increase the pressure and temperature of steam and reduce the loss in the energy conversion process, so that the unit can achieve higher efficiency in the process of converting thermal energy into electrical energy. Taking ultra-supercritical units as an example, compared with traditional subcritical units, their power generation efficiency can be increased by about 6%-10%, which is effective in reducing energy consumption and carbon emissions.

[0004] As the core equipment of the generator set, the steam turbine plays a vital role in the entire power generation process. It converts the thermal energy of steam into mechanical energy and then drives the generator to generate electricity. The stability and reliability of its operation are directly related to the overall operating efficiency and economic benefits of the power plant. However, with the increase of unit capacity and the improvement of operating parameters, the operating environment faced by the steam turbine has become more complex, and the risk of failure has also increased significantly.

[0005] From the perspective of mechanical structure, the size and weight of large-capacity steam turbines have increased significantly, and each component is subjected to tremendous pressure and temperature stress under the action of high-parameter steam. For example, the high-pressure rotor needs to operate for a long time under conditions of high temperature, high pressure and high-speed rotation, which places extremely high demands on the strength, toughness and high-temperature resistance of its materials. Once the material suffers fatigue damage or performance degradation, it may cause serious accidents such as rotor cracks and fractures. Among steam turbine failures, the downtime caused by rotor failure accounts for a large proportion, which not only causes a large loss of power generation, but also causes a sharp increase in maintenance costs.

[0006] In terms of control systems, high-parameter operation requires higher control accuracy and faster response speed for steam turbines. Although advanced digital control systems (DCS) have improved control performance to a certain extent, they have also introduced new risk factors. Software failures, communication failures or logical errors in the DCS system may cause abnormal control of the steam turbine, leading to abnormal operation of the unit or even shutdown accidents. Moreover, as the degree of automation of modern steam turbines increases, the dependence of operators on the control system also increases accordingly. Once a problem occurs in the control system, it becomes more difficult for operators to accurately judge and handle the fault in a short period of time.

[0007] In addition, the operation of steam turbines is also affected by a variety of external factors. For example, fluctuations in coal quality, changes in water quality, and frequent adjustments in power grid loads may have an adverse effect on the normal operation of steam turbines. Unstable coal quality will lead to a decline in steam quality, causing scaling and corrosion of turbine blades, and reducing the efficiency and reliability of steam turbines; poor water quality may cause corrosion and scaling of internal parts of steam turbines, affecting the service life of the equipment.

[0008] The economic losses caused to power plants by steam turbine accidents cannot be underestimated. Minor faults may cause the unit to trip for a short time, resulting in the loss of several hours of power generation during the process of reconnecting to the grid after the fault is eliminated. Serious faults will not only damage steam turbine components, requiring a large amount of money to repair or replace, but may also cause long-term downtime, causing the power plant to lose huge amounts of power generation revenue. According to industry data, a serious steam turbine failure may result in direct economic losses of millions or even tens of millions of yuan, and indirect economic losses are even more difficult to estimate, including the impact on grid stability, reduced reliability of power supply to users, and damage to corporate reputation.

[0009] In summary, in the context of power plants upgrading to large capacity and high parameters, the operational safety of steam turbines faces unprecedented challenges. Therefore, while improving the power generation efficiency of the units, strengthening the troubleshooting research on steam turbines and improving their protection measures have become important issues that need to be urgently addressed in the power generation industry; it is not only related to the economic benefits and operational stability of power plants, but also has far-reaching significance for ensuring the reliability and safety of power supply. Summary of the invention

[0010] In view of the problems existing in the prior art, the present invention provides a steam turbine fault troubleshooting method which is based on the statistical results of past big data cases and is accurately classified and executed.

[0011] To achieve the above purpose, the technical solution adopted by the present invention is as follows: The present invention provides a steam turbine troubleshooting method based on big data, which comprises the following steps: Collecting historical fault information of the steam turbine, and pre-classifying the fault type based on the industrial control information in the historical fault information; When the result of the pre-classification is an industrial control fault, outputting an industrial control fault correction strategy through a multi-level analysis and judgment method; When the result of the pre-classification is a non-industrial control fault, a priority sequence of typical faults in a big data statistical case is calculated based on a multi-criteria decision evaluation algorithm; A troubleshooting strategy based on the priority sequence is established through a mapping method.

[0012] Optionally, the calculation steps of the multi-criteria decision-making evaluation algorithm include: Building a decision matrix based on the information of the typical faults; Standardizing the elements of the decision matrix to obtain a standardized decision matrix; The standardized decision matrix is ​​weighted by an entropy weight method to obtain a weighted decision matrix; The priority of each typical fault is calculated based on the topsis method, and the typical faults are sorted according to the priority to obtain the priority sequence.

[0013] Optionally, the evaluation criteria in the decision matrix include probability of occurrence, degree of impact, difficulty of repair and risk of downtime; The results of the evaluation criteria are quantitative analysis results.

[0014] Optionally, the result of the evaluation criterion is an expert system score or a quantitative calculation result of historical data.

[0015] Optionally, in the standardization process of the elements of the decision matrix, the range method is used for processing, and the evaluation criteria are converted into efficiency indicators to map the original data of the evaluation criteria to the interval of 0-1.

[0016] Optionally, in weighting the standardized decision matrix by an entropy weight method, the proportion, information entropy and difference coefficient of the elements in the standardized decision matrix are calculated in sequence, and the entropy weight is obtained; The normalized decision matrix is ​​weighted by the entropy weight.

[0017] Optionally, the typical faults include transmission system fault, lubrication system fault, valve fault, blade fault and control system communication fault.

[0018] Optionally, the industrial control failure includes an industrial control system operation failure, an industrial control system logic design failure, and an industrial control system operation failure; The non-industrial control failures include equipment damage, on-site misoperation and the impact of human behavior.

[0019] Optionally, the industrial control fault correction strategy includes restarting the industrial control system or starting a backup industrial control system.

[0020] Optionally, the troubleshooting strategy is a troubleshooting order of the typical faults based on a priority sequence and a corresponding optimal troubleshooting strategy; the step of obtaining the optimal troubleshooting strategy includes: Collect past troubleshooting strategies and corresponding total failure losses under the typical failure conditions; The elimination strategy corresponding to the smallest total failure loss is selected as the optimal elimination strategy.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention divides steam turbine faults by taking faults caused by industrial control as the starting point, pre-analyzes the fault types from the equipment layer and human factors, and prioritizes them in combination with typical fault cases of historical steam turbines, so as to sequentially execute corresponding troubleshooting strategies to reduce fault losses and provide an execution sequence reference for troubleshooting operations under conditions of multiple concurrent faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be 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 labor.

[0023] Figure 1 It is a method flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0026] It is worth noting that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products, and their sources are not specifically limited unless otherwise specified.

[0027] It should also be noted that in order to facilitate understanding in the specific embodiments of the present invention, the method steps are described in a certain order, but those skilled in the art can change the order of the steps according to actual needs, and therefore this cannot be used as a limiting condition; it should be further noted that in the description of the following specific embodiments, the upper and lower subscripts of each parameter should be understood as distinguishing marks of similar identifications with reference to common interpretations unless otherwise specified, representing the parameters of the device related or corresponding to the subscript, and should not be understood as specific models or special marks.

[0028] See also Figure 1 As shown, a steam turbine troubleshooting method based on big data is used to generate a troubleshooting template, which is then used in the current fault processing process to improve troubleshooting efficiency and reduce fault losses; it includes the following steps: Collect and compile historical fault information of steam turbines, and pre-classify fault types based on the industrial control information in the historical fault information; Specifically, in the pre-classification process, the fault information is differentiated and compared, and the industrial control information and non-industrial control information are compared for relevance. When the industrial control information is abnormal and the non-industrial control information related to the industrial control information is normal (such as the valve control signal is abnormal, while the valve opening sensor and upstream and downstream flow sensor detection results show normal), or the non-industrial control information related to the industrial control information is controlled abnormal, it means that the fault type is an industrial control fault. Among them, industrial control faults include industrial control system operation faults, industrial control system logic design faults, and industrial control system operation faults; when the accident is an industrial control fault, that is, the equipment itself is operating normally, but there are faults at the control level due to poor control, improper operation, and other factors that affect the operation of the unit. Industrial control fault correction strategies include but are not limited to restarting the industrial control system or starting a backup industrial control system.

[0029] Therefore, a multi-level analysis and judgment method is used to judge and output the corresponding solution strategy (i.e., industrial control fault correction strategy); First, check whether it is a DCS system failure based on the abnormal information content (including DCS system shutdown, DPU damage, etc., which cause the unit to lose control and are classified as DCS system failure). If so, repair the DCS, restart the DCS or enable the backup. If not, the next step of analysis is to trace back the abnormal information according to the system operation log, and perform logical deduction on the abnormal information to determine whether it is due to improper logic design (including logical contradictions, lack of corresponding logic leading to poor unit control, etc., which are classified as improper logic failures). If so, the backup DCS is activated and the main DCS is upgraded; If not, the next step of analysis is to trace back the personnel operation process based on the monitoring information, and compare the personnel operation process with a reference (reference is the operation manual) to determine whether it is caused by improper operation of the operator (including operators not operating according to the operating procedures, improper water level control, etc., which affect the normal operation or even damage of the turbine and are classified as improper operation failures). If so, roll back the DCS; If not, the next step of analysis is carried out, that is, the system alarm records are traced back according to the monitoring information, and the alarm duration is compared with the time threshold. When the time threshold is not exceeded, the DCS is rolled back. Such faults are usually discovered in time to avoid machine tripping or more serious faults, but when the time threshold is exceeded, the automatic machine trip protection is activated. If the vibration value rises but is not discovered in time, the preset automatic machine trip protection will trigger the automatic shutdown or automatically notify the local team to destroy the vacuum to reduce the losses caused by equipment damage; if the unit can be restored to normal operation through automatic measures, such as designing automated logic to control valves to modify water levels, it can be restored in time to reduce the economic losses caused by shutdown, or the local team can be notified to perform local operations.

[0030] When the result of the pre-classification is an industrial control failure, the industrial control failure correction strategy is initiated; among them, non-industrial control failures include equipment damage, on-site misoperation and the impact of personnel behavior, which are characterized by normal industrial control information and abnormal non-industrial control information related to the industrial control information.

[0031] When the result of pre-classification is a non-industrial control fault, the priority sequence of typical faults in the big data statistical case is first calculated based on the multi-criteria decision evaluation algorithm. Among them, the calculation steps of the multi-criteria decision-making evaluation algorithm include: A decision matrix is ​​constructed based on information of typical faults; optionally, typical faults include transmission system faults, lubrication system faults, valve faults, blade faults and control system communication faults; evaluation criteria in the decision matrix include probability of occurrence, degree of impact, difficulty of repair and risk of downtime; wherein the results of the evaluation criteria are quantitative analysis results.

[0032] Therefore, in this embodiment, some common typical fault composition examples are selected, such as: Excessive bearing vibration 1#; Lubricating oil system failure 2#; Steam valve stuck 3#; Blade scaling or corrosion 4#; Control system communication failure 5#; To assess the risk of these failures, the following assessment criteria were selected: Probability of occurrence (P): the possibility of a fault occurring (1-10 points, 10 points is the highest); Impact degree (I): the impact of the fault on the operation of the equipment (1-10 points, 10 points is the highest); Repair difficulty (D): the complexity and time cost of fault repair (1-10 points, 10 points is the highest); Downtime risk (S): The probability that a failure will lead to downtime (1-10 points, 10 is the highest).

[0033] In order to clearly quantify the above evaluation criteria, the results are expert system scores or historical data quantitative calculation results. The expert system scores based on past data using a preset algorithm, or it can use the fault scoring system in the existing technology to score, and then scale the data according to the calculation requirements (in the range of 1-10). The quantitative calculation method of historical data is to perform global statistics and calculations in a normalized manner, with an interval of 1-10.

[0034] The elements of the decision matrix are standardized to obtain a standardized decision matrix; optionally, the elements of the decision matrix are standardized by using the range method, and the evaluation criteria are converted into benefit indicators (the larger the value, the higher the risk), so as to map the original data of the evaluation criteria to the interval of 0-1, and the formula is: ; In the formula, r ij : No. i The scheme is in j The original value under the criteria; min i r ij : No. j The minimum value of all solutions under the criteria; max i r ij : No. j The maximum value of all solutions under the criteria; x ij : No. i The scheme is in j Standardized value under this criterion.

[0035] Therefore, the standardized decision matrix is: The standardized decision matrix is ​​weighted by the entropy weight method, and a weighted decision matrix is ​​obtained; further, in the process of weighting the standardized decision matrix by the entropy weight method, the proportion of elements in the standardized decision matrix, information entropy and difference coefficient are calculated in turn, and the entropy weight is obtained; the standardized decision matrix is ​​weighted by the entropy weight.

[0036] Calculate entropy value and entropy weight; (1) Calculation of the proportion of indicators: ; In the formula, pij : No. i The scheme is in j The proportion under each criterion; m : The total number of solutions.

[0037] (2) Calculate information entropy: ; In the formula, e j : No. j The information entropy of a criterion.

[0038] (3) Calculate the coefficient of variation: ; In the formula, d j : No. j The coefficient of variation of the criteria.

[0039] (4) Determine entropy weight: ; In the formula, w j : No. j The entropy weight of a criterion; n : Total number of criteria.

[0040] Further, the demonstration calculation results of this embodiment are: Entropy e j : [0.95, 0.92, 0.89, 0.91]; Coefficient of variation d j : [0.05, 0.08, 0.11, 0.09]; Entropy Weight w j : [0.15, 0.24, 0.33, 0.28]; Going a step further, we can weight the normalized decision matrix and combine the normalized matrix with the entropy weight: (Where, v ij : No. i The scheme is in j The entropy weight of the weighted standardized value under each criterion is: The priority of each typical fault is calculated based on the topsis method, and the typical faults are sorted by priority to obtain a priority sequence.

[0041] First, determine the positive / negative ideal solution in the topsis method calculation Positive ideal solution ; In the formula, A + : Positive ideal solution, indicating the optimal value under each criterion; max i v ij : No. j The maximum value of all solutions under the criteria.

[0042] Calculation results: A + : [0.15, 0.24, 0.33, 0.28]; Negative ideal solution ; In the formula, A — : Positive ideal solution, indicating the worst value under each criterion; min i v ij : No. j The minimum value of all solutions under the criteria.

[0043] Calculation results: A − : [0.05, 0.08, 0.11, 0.09].

[0044] Second, distance and relative proximity (i.e., similarity) are calculated; 1. Distance to the positive ideal solution: ; In the formula, D i + : No. i The Euclidean distance from a solution to the positive ideal solution.

[0045] 2. Distance to negative ideal solution: ; In the formula, D i — : No. i The Euclidean distance from a solution to the positive ideal solution.

[0046] 3. Relative proximity: ; In the formula, C i : No. i The relative closeness of the two solutions. The larger the value, the better the solution.

[0047] Therefore, the calculation results are updated to show: Sort by relative proximity Ci: 1. Control system communication failure (Ci=1.00); 2. The bearing vibration is too large (Ci=0.60); 3. Blade fouling or corrosion (Ci=0.40); 4. Lubricating oil system failure (Ci=0.33); 5. Steam valve stuck (Ci=0.27); For different steam turbine units, the priority sequence of typical faults among non-industrial control faults is calculated in advance according to the method of this embodiment. When there are multiple tasks to be processed in the same period during daily operation (such as when multiple non-industrial control faults occur), they are processed according to the solved priority sequence and the corresponding troubleshooting strategy is output. Furthermore, the corresponding troubleshooting strategy includes the troubleshooting order of typical faults based on the priority sequence and the corresponding optimal troubleshooting strategy. Among them, the steps of obtaining the optimal troubleshooting strategy include: Collect past troubleshooting strategies and corresponding total failure losses under typical failure conditions; the total failure loss calculation method includes: First, determine the indirect losses: Loss from production interruption: If the fault causes the turbine to shut down, it will cause the power plant to generate less electricity. The amount of electricity that is less generated can be calculated based on the power generation capacity of the power plant and the time to troubleshoot the fault. Combined with the profit per kilowatt-hour, the economic loss caused by the production interruption can be obtained. For example, if the power generation capacity of the power plant is P megawatts, the troubleshooting time is t hours, and the profit per kilowatt-hour is m, the loss from production interruption = P×t×1000×m (convert megawatts to kilowatts).

[0048] Subsequent losses due to equipment damage: In addition to the cost of repairing equipment damage caused directly by the failure (i.e. part of the direct loss), it is also necessary to consider the long-term losses that may be caused by the performance degradation of the equipment after repair, such as increased energy consumption and shortened service life of the equipment after repair. Assuming that the annual energy cost of the equipment after repair is C1, and the remaining service life is shortened by n years, this part of the loss can be expressed as C1×n. At the same time, if the equipment is scrapped prematurely and needs to be replaced with a new one, the difference between the purchase and installation cost of the new equipment and the expected income of the old equipment within the remaining service life should also be included in the indirect loss.

[0049] Second, consider the opportunity cost: Missed power generation revenue opportunities: Due to fault shutdown, the power plant may miss some power generation opportunities during high electricity price periods. By analyzing historical electricity price data, the additional revenue that could be obtained if normal power generation was performed during the fault period is calculated as part of the opportunity cost and set as C2.

[0050] Finally, calculate the total loss of the fault: Total loss of the fault = direct loss of the fault (cost of damaged equipment) + production interruption loss + subsequent loss of equipment damage + missed power generation revenue opportunity, that is, comprehensive loss = direct loss of the fault + P×t×1000×m+C1×n+C2. In this way, the various impacts of the fault are fully considered, the comprehensive loss can be calculated more accurately, and the elimination strategy corresponding to the smallest total loss of the fault can be selected from the historical data, and a mapping relationship is established between it and the typical fault, which is recorded as the optimal elimination strategy under the typical fault condition.

[0051] The troubleshooting sequence and its optimal troubleshooting strategy together constitute the troubleshooting strategy.

[0052] Current troubleshooting process: After the system detects fault information, it performs part of the process of the above template, that is, first pre-classification is performed. If the result is an industrial control fault, the corresponding solution strategy can be selected according to the multi-level analysis and judgment method; if the result is a non-industrial control fault, a matching typical fault is selected according to the fault characteristics, and its priority is queried. In the multi-task concurrent scenario, troubleshooting is performed according to the order in the priority sequence, and the corresponding optimal troubleshooting strategy is executed. Among them, it should be noted that in the multi-task concurrent scenario, it is executed according to the resource situation (personnel, tools, parts). When the resources meet the simultaneous execution of all tasks, troubleshooting can be performed simultaneously without following the order of the priority sequence; when only some tasks are supported to be performed simultaneously, troubleshooting tasks with higher troubleshooting priorities can be performed simultaneously.

[0053] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A steam turbine troubleshooting method based on big data, characterized in that: The steps include: Collecting historical fault information of the steam turbine, and pre-classifying the fault type based on the industrial control information in the historical fault information; When the result of the pre-classification is an industrial control fault, outputting an industrial control fault correction strategy through a multi-level analysis and judgment method; When the result of the pre-classification is a non-industrial control fault, a priority sequence of typical faults in a big data statistical case is calculated based on a multi-criteria decision evaluation algorithm; The calculation steps of the multi-criteria decision-making evaluation algorithm are: Building a decision matrix based on the information of the typical faults; Standardizing the elements of the decision matrix to obtain a standardized decision matrix; The standardized decision matrix is ​​weighted by an entropy weight method to obtain a weighted decision matrix; Calculate the priority of each typical fault based on the topsis method, and sort the typical faults according to the priority to obtain the priority sequence; A troubleshooting strategy based on the priority sequence is established through a mapping method.

2. The method for troubleshooting a steam turbine based on big data according to claim 1, characterized in that: The evaluation criteria in the decision matrix include probability of occurrence, degree of impact, difficulty of repair and risk of downtime; The results of the evaluation criteria are quantitative analysis results.

3. The method for troubleshooting a steam turbine based on big data according to claim 2, characterized in that: The result of the evaluation criteria is an expert system score or a quantitative calculation result of historical data.

4. The method for troubleshooting a steam turbine based on big data according to claim 1, 2 or 3, characterized in that: In the process of standardizing the elements of the decision matrix, the range method is used for processing, and the evaluation criteria are converted into efficiency indicators to map the original data of the evaluation criteria to the interval of 0-1.

5. The method for troubleshooting a steam turbine based on big data according to claim 1, characterized in that: In weighting the standardized decision matrix by the entropy weight method, the proportion, information entropy and difference coefficient of the elements in the standardized decision matrix are calculated in sequence, and the entropy weight is obtained; The normalized decision matrix is ​​weighted by the entropy weight.

6. The method for steam turbine troubleshooting based on big data according to claim 1, characterized in that: The typical faults include transmission system fault, lubrication system fault, valve fault, blade fault and control system communication fault.

7. The method for troubleshooting a steam turbine based on big data according to claim 1, characterized in that: Industrial control failures include industrial control system operation failures, industrial control system logic design failures, and industrial control system operation failures; The non-industrial control failures include equipment damage, on-site misoperation and the impact of human behavior.

8. The method for troubleshooting a steam turbine based on big data according to claim 7 is characterized in that: The industrial control fault correction strategy includes restarting the industrial control system or starting a backup industrial control system.

9. The method for steam turbine troubleshooting based on big data according to claim 1, characterized in that: The troubleshooting strategy is the troubleshooting order of the typical faults based on the priority sequence and the corresponding optimal troubleshooting strategy; The step of obtaining the optimal exclusion strategy includes: Collect past troubleshooting strategies and corresponding total failure losses under the typical failure conditions; The elimination strategy corresponding to the smallest total failure loss is selected as the optimal elimination strategy.

Images