Power plant equipment maintenance quality evaluation method

Abstract

The application provides a power station equipment maintenance quality evaluation method, comprising the following steps: 1) constructing an initial state model S based on historical operation data after the last maintenance; combining operation condition parameters after the maintenance and the initial state model S to calculate full-condition operation state prediction values S(1, 2,..., N); 2) constructing a pre-maintenance state model E based on historical operation data before the maintenance; combining operation condition parameters after the maintenance and the pre-maintenance state model E to calculate full-condition operation state prediction values E(1, 2,..., N); 3) calculating maintenance scores of the operation state under each condition according to the deviation between actual values T(1, 2,..., N) and prediction values S(1, 2,..., N), E(1, 2,..., N) of the operation state after the maintenance; 4) obtaining full-condition maintenance scores of the operation state by weighting; 5) obtaining maintenance scores of the equipment by combining operation state weights and the full-condition maintenance scores of each operation state, and completing the maintenance quality evaluation. The application comprehensively considers the deviation of the operation state before and after the maintenance under different conditions, and provides a quantitative evaluation method for the maintenance quality of the power station equipment.

Description

Technical Field

[0001] This invention relates to the field of power plant equipment maintenance and evaluation, and specifically to a method for evaluating the quality of power plant equipment maintenance. Background Technology

[0002] Hydropower equipment, especially pumped-storage units, is facing increasingly frequent start-ups, shutdowns, and operational mode transitions. Under the influence of hydraulic, mechanical, and electrical factors, it gradually deteriorates. Factors such as hydraulic shock, mechanical failure, and electromagnetic imbalance can also induce various faults and accidents in hydropower equipment, including abnormal vibration, structural fatigue, electrical faults, and disruptions to operating modes. Unit overhaul, as a maintenance method to restore equipment function and performance, is widely used in hydropower stations and pumped-storage power stations, especially in the middle and later stages of the bathtub failure curve, where the equipment failure rate gradually increases. While overhaul cannot restore the unit to its optimal state, it can effectively optimize the equipment condition and reduce potential failure rates. Based on the scale of unit overhaul and the duration of downtime, unit overhaul is divided into four levels: A, B, C, and D. Level A overhaul refers to a comprehensive disassembly inspection and repair of the generator unit to maintain, restore, or improve equipment performance. Level B overhaul refers to the disassembly inspection and repair of certain equipment within the unit, addressing problems with specific components. Level C maintenance refers to the focused inspection, evaluation, repair, and cleaning of the unit based on the wear and aging patterns of the equipment. Level D maintenance refers to the troubleshooting of auxiliary systems and equipment of the main equipment when the overall operating condition of the unit is good.

[0003] Currently, there are relevant standards and specifications for power plant unit maintenance, and each power plant has also developed its own maintenance procedures to guide maintenance work. However, as a routine means of maintaining and restoring unit condition and performance, maintenance quality is generally determined through expert acceptance and evaluation, lacking available quantitative assessment methods and means for unit maintenance quality. Vibration and sway are characteristic operating states reflecting the health status of power plant equipment, affected by component manufacturing and installation quality and operating conditions, and are characteristic indicators reflecting unit maintenance quality. Currently, unit maintenance quality assessment focuses on comparison with before maintenance, only indicating whether the condition is better after maintenance than before, but unable to quantitatively assess the degree of restoration of unit function and performance.

[0004] Chinese patent document CN 112633534A discloses a comprehensive evaluation method and system for the maintenance effect of pumped storage units. It obtains maintenance quality evaluation results by comparing the degree of deterioration before and after maintenance. However, this method suffers from several drawbacks. First, it is difficult to define and quantify the degree of deterioration, requiring the introduction of numerous custom thresholds, which can lead to significant subjective influences. Second, it lacks a comparative benchmark, making it difficult to quantify the unit's recovery status after maintenance. Chinese patent document CN 116911820A discloses a digital holographic control system for the maintenance of hydropower station units. This system follows the method of CN 112633534A, using VR-based real-scene reconstruction for maintenance simulation to evaluate maintenance quality, and therefore suffers from the same shortcomings. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies. Taking into account the characteristics of the bathtub failure curve of power plant equipment—high at both ends, low in the middle, and continuous throughout—this invention introduces the state that the equipment can recover to after the last maintenance as the initial benchmark, considering the equipment's entire life cycle. Furthermore, it quantitatively evaluates the quality of the current maintenance by combining the equipment's operating status before and after the previous maintenance with that of the current maintenance. To achieve the above objective, this invention is implemented through the following technical solution:

[0006] A method for evaluating the quality of power plant equipment maintenance includes:

[0007] Step S1: Construct an initial state model S based on historical operating data after the last overhaul; Combine the operating condition parameters after this overhaul with the initial state model S to calculate the initial predicted value S(1,2,…,N) of the full operating condition;

[0008] Step S2: Construct a pre-maintenance state model E based on historical operating data before this maintenance; combine the operating condition parameters after this maintenance and the pre-maintenance state model E to calculate the pre-maintenance predicted value E(1,2,…,N) of the full operating condition;

[0009] Step S3: Calculate the maintenance score for each operating condition after maintenance based on the deviation between the actual value T(1,2,…,N) and the predicted values ​​S(1,2,…,N) and E(1,2,…,N) of the operating condition after this maintenance.

[0010] Step S4: Combine the maintenance scores of the operating state under each operating condition after maintenance, and obtain the weighted full-condition maintenance score of the operating state.

[0011] Step S5: Combine the weight of the operating status and the full-condition maintenance score of each operating status to obtain the maintenance score of the equipment and complete the maintenance quality evaluation.

[0012] The operating data mentioned in steps S1 and S2 includes operating parameters and operating status. The operating parameters include active power, reactive power, head, and flow rate. The operating status is the vibration and sway of the equipment. Steps S1 and S2 are not sequential.

[0013] Further, step S1 includes:

[0014] Step S11: Collect historical operating data of the equipment after the last maintenance, including operating parameters such as active power P, reactive power Q, head h, flow rate q, and equipment operating status such as vibration v and swing s.

[0015] Step S12: Based on the long short-term memory neural network, learn and train the historical operating data of the equipment to obtain the vibration initial state model S of the equipment. v = f(P,Q,h,q) and the initial state model of the swing S s =f(P,Q,h,q), where the initial state model reflects the mapping relationship between the equipment's operating state and operating parameters after the last overhaul.

[0016] Further, step S2 includes:

[0017] Step S21: Collect historical operating data of the equipment before this overhaul, including operating parameters such as active power P, reactive power Q, head h, flow rate q, and equipment operating status such as vibration v and swing s.

[0018] Step S22: Based on the long short-term memory neural network, learn and train the historical operating data of the equipment to obtain the pre-vibration repair state model E of the equipment. v = f(P,Q,h,q) and the slew rate pre-correction state model E s =f(P,Q,h,q), where the pre-maintenance state model reflects the mapping relationship between the equipment's operating status and working parameters before this maintenance.

[0019] Furthermore, step S3 uses piecewise interpolation to calculate the maintenance score for each operating condition after this maintenance, and the calculation formula is as follows:

[0020]

[0021] Where D(i) is the maintenance score of the operating status under the i-th operating condition, T(i) is the actual value of the operating status under the i-th operating condition, S(i) is the initial predicted value of the operating status under the i-th operating condition, and E(i) is the pre-repair predicted value of the operating status under the i-th operating condition.

[0022] Furthermore, step S4 uses the averaging method to obtain the full-condition maintenance score D for each operating state, and the calculation formula is as follows:

[0023]

[0024] Furthermore, the operating state weight p mentioned in step S5 is determined using the three-scale method, and the equipment maintenance score DT is calculated using the following formula:

[0025]

[0026] Where, p i Let D be the weight of the i-th running state. i Let m be the maintenance score for the i-th operating state, and m be the number of operating states.

[0027] Furthermore, the three-scale method for determining state weights includes:

[0028] Step S51: Construct a three-scale judgment matrix for each running state based on the relative importance of each running state;

[0029] Step S52: Calculate the weights of each operating state to obtain the decision weight feature vector of each operating state;

[0030] Step S53: Perform a consistency check on the weight feature vectors of each operating state;

[0031] Step S54: If the consistency check meets the standard, determine the weight of each operating state; if it does not meet the standard, adjust the judgment matrix and repeat steps S51 to S73.

[0032] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0033] This invention combines the characteristics of the bathtub failure curve of power plant equipment, which is high at both ends and low in the middle and continuous before and after. It takes the previous maintenance as the initial benchmark for the current maintenance and the data before the current maintenance as the basis for comparison. On the one hand, it evaluates whether the current maintenance is better than the previous maintenance, and on the other hand, it evaluates whether the current maintenance has restored the state after the previous maintenance. This realizes the quantitative evaluation of maintenance quality based on the characteristics of the equipment's full life cycle failure bathtub curve. Attached Figure Description

[0034] Figure 1 This is a flowchart of a method for evaluating the maintenance quality of power plant equipment according to the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the technical solutions of the present invention, preferred embodiments of the present invention are described below in conjunction with specific examples. However, it should be understood that the accompanying drawings are for illustrative purposes only and should not be construed as limiting the present invention. For better illustration of this embodiment, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable that some well-known structures and their descriptions may be omitted in the drawings for those skilled in the art. The positional relationships described in the drawings are for illustrative purposes only and should not be construed as limiting the present invention.

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the present invention.

[0037] like Figure 1 The method for evaluating the quality of power plant equipment maintenance, taking the maintenance evaluation of Unit 1 of a pumped storage power station as an example, includes:

[0038] Step S1: Collect historical operating data of Unit 1 for one week after the last overhaul, including operating parameters such as active power P, reactive power Q, head h, and flow rate q, as well as 15 operating states such as vibration and sway (see Table 1 for details); divide the one-week data into 5 days of training data and 2 days of test data; use the 5 days of training data to learn and train based on a long short-term memory neural network to obtain the initial state models S1 to S2 for each operating state of the equipment. 15 The initial state model is expressed as S = f(P, Q, h, q), which reflects the mapping relationship between the equipment operating status and operating parameters after the last overhaul; the trained model is tested using the remaining 2 days of test data.

[0039] Table 1 Unit Operating Status

[0040]

[0041] Based on the actual values ​​of operating parameters such as active power, reactive power, flow rate, and head during the first week after this overhaul and the initial state model S1~S 15 Calculate the initial predicted values ​​S1(1,2,…,N)~S for 15 operating states under actual operating conditions. 15 (1,2,…,N).

[0042] Step S2: Collect historical operating data of Unit 1 for one week prior to this overhaul, including operating parameters such as active power P, reactive power Q, head h, and flow rate q, as well as 15 operating states such as vibration and sway. Divide the week's data into 5 days of training data and 2 days of test data. Use the 5 days of training data to learn and train based on a long short-term memory neural network to obtain the pre-overhaul state models E1 to E2 for each operating state of the equipment. 15The pre-maintenance state model is expressed as E = f(P, Q, h, q), which reflects the mapping relationship between the equipment operating status and operating parameters before this maintenance; the trained model is tested using the remaining 2 days of test data.

[0043] Based on the actual values ​​of active power, reactive power, flow rate, and head, and the pre-maintenance state models E1 to E1, this maintenance was conducted for one week after the maintenance. 15 Calculate the pre-repair predicted values ​​E1(1,2,…,N)~E under 15 operating conditions in actual operation. 15 (1,2,…,N).

[0044] Step S3: Based on the actual operating status values ​​T1(1,2,…,N)~T after one week of operation following this overhaul... 16 (1,2,…,N), and the initial predicted values ​​S1(1,2,…,N)~S 15 (1,2,…,N), Pre-revision predicted value E1(1,2,…,N)~E 15 The deviation of (1,2,…,N) is used to calculate the maintenance score for each operating condition after maintenance. If the actual value is not worse than the corresponding initial predicted value, the maintenance objective has been achieved, and the maintenance score is 100 points. If the actual value is worse than the pre-maintenance predicted value, the maintenance has not achieved any state restoration effect on the equipment, and the maintenance score is 0 points. If the actual value is between two predicted values, the maintenance score is calculated by interpolation. Taking the maintenance score calculation of the guide bearing + X-direction runout as an example, the specific calculation formula is as follows:

[0045]

[0046] Wherein, D1(i) is the maintenance score of the upper guide bearing +X direction runout under the i-th operating condition, T1(i) is the actual value of the upper guide bearing +X direction runout under the i-th operating condition, S1(i) is the initial predicted value of the upper guide bearing +X direction runout under the i-th operating condition, and E1(i) is the pre-repair predicted value of the upper guide bearing +X direction runout under the i-th operating condition.

[0047] Step S4: Combining the maintenance scores of each operating condition after maintenance, a weighted maintenance score for the entire operating condition is obtained. Since a corresponding maintenance score is obtained for each actual operating condition, the overall maintenance score for the operating condition is calculated using the averaging method. The calculation of the overall maintenance score for the upper guide bearing + X-direction runout is as follows:

[0048]

[0049] Optionally, considering the differences between long-term operating conditions and the data collected during the first week of operation, the operating history from the last overhaul to the current overhaul can be used as the basis for weight calculation. The frequency of occurrence of all operating conditions during the statistical period is calculated, and the weight of each operating condition in the current week of operation is obtained through normalization calculation. Finally, the weighted overhaul score of the entire operating condition of the upper guide bearing + X-axis swing is obtained.

[0050] Step S5: Addressing the issues of excessive scaling in the commonly used nine-scale method, inconsistent understanding of scaling among different individuals, poor accuracy in comparison scaling, and inconsistent comparison matrices, this solution employs a three-scale method to determine the weights p1 to p2 for the 15 operating states. 15 Based on the above calculations, the full-condition maintenance scores D1 to D2 for the 15 operating states are obtained. 15 The maintenance score DT for Unit 1 during this overhaul is obtained using the following formula:

[0051]

[0052] The three-scale method for determining state weights includes:

[0053] Step S51: Construct the judgment matrix: Based on the relative importance among the running states, construct a three-scale comparison matrix M for the running states. The importance judgment is as follows:

[0054]

[0055] The sum of the elements in each row of the comparison matrix M is calculated as follows:

[0056]

[0057] r i The maximum value r in max and minimum value r min The two corresponding indicators are compared according to the nine-scale 1-9 to obtain the baseline comparison scale b. m Finally, the direct three-scale comparison matrix is ​​transformed into an indirect judgment matrix C = (c ij n×n:

[0058]

[0059] Step S52: Calculate the running state weights: After obtaining the judgment matrix C, the running states X1, X2, ..., X can be obtained. 15 The weights ω1, ω2, ..., ω relative to Unit 1 15 This can be written in vector form as W = (ω1, ω1, ..., ω1). T .

[0060] From the equation

[0061] CW=λ max W

[0062] Find the largest eigenvalue λ of the judgment matrix. max And the corresponding eigenvectors W, which, after normalization, become the index states X1, X2, ..., X. 15 Weight A of Unit 1 X ;

[0063] Step S53: Perform a consistency check: Calculate the consistency index CI.

[0064]

[0065] The corresponding average random consistency index RI is obtained by looking up the table:

[0066] order 1 2 3 4 5 6 7 8 9 RI 0 0 0.58 0.9 1.12 1.24 1.32 1.41 1.45

[0067] Calculate the consistency ratio (CR):

[0068]

[0069] Step S54: When CR < 0.1, the judgment matrix is ​​considered to meet the consistency requirements and is acceptable. The running state weight is A. X When CR > 0.1, it is considered that the judgment matrix does not meet the consistency requirements and needs to be revised. The judgment matrix is ​​then adjusted, and steps S51 to S54 are repeated.

[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for evaluating the quality of power plant equipment maintenance, characterized in that, include: Step S1: Construct an initial state model S based on historical operating data since the last maintenance; Based on the operating condition parameters after this overhaul and the initial state model S, calculate the initial predicted value S(i) of the full operating condition, i=1,2,…,N; Step S2: Construct a pre-maintenance state model E based on historical operating data before this maintenance; combine the operating condition parameters after this maintenance and the pre-maintenance state model E to calculate the pre-maintenance predicted value E(i) of the full operating condition, i = 1, 2, ..., N; Step S3: Calculate the maintenance score D(i), i = 1, 2, ..., N, based on the deviation between the actual value T(i), i = 1, 2, ..., N, the initial predicted value S(i), i = 1, 2, ..., N, and the pre-maintenance predicted value E(i), i = 1, 2, ..., N, for each operating condition after maintenance. Step S4: Combine the maintenance scores of the operating state under each operating condition after maintenance, and obtain the weighted full-condition maintenance score of the operating state. Step S5: Combine the weight of the operating status and the full-condition maintenance score of each operating status to obtain the maintenance score of the equipment and complete the maintenance quality evaluation. The operating data mentioned in steps S1 and S2 includes operating parameters and operating status. The operating parameters include active power, reactive power, head, and flow rate. The operating status is the vibration and sway of the equipment. Steps S1 and S2 are not in any particular order. In step S3, the piecewise interpolation method is used to calculate the maintenance score for each operating condition after this maintenance. The calculation formula is as follows: Where D(i) is the maintenance score of the operating status under the i-th operating condition, T(i) is the actual value of the operating status under the i-th operating condition, S(i) is the initial predicted value of the operating status under the i-th operating condition, and E(i) is the pre-repair predicted value of the operating status under the i-th operating condition. In step S4, the average method is used to obtain the full-condition maintenance score D for each operating state. The calculation formula is as follows:

2. The method for evaluating the maintenance quality of power plant equipment according to claim 1, characterized in that, Step S1 includes: Step S11: Collect historical operating data of the equipment after the last overhaul, including operating parameters and equipment operating status. The operating parameters include active power P, reactive power Q, head h, and flow rate q. The equipment operating status includes vibration v and swing s. Step S12: Based on the long short-term memory neural network, learn and train the historical operating data of the equipment to obtain the vibration initial state model S of the equipment. v = f(P,Q,h,q) and the initial state model of the swing S s =f(P,Q,h,q), where the initial state model reflects the mapping relationship between the equipment's operating state and operating parameters after the last overhaul.

3. The method for evaluating the maintenance quality of power plant equipment according to claim 1, characterized in that, Step S2 includes: Step S21: Collect historical operating data of the equipment before this overhaul, including operating parameters and equipment operating status. The operating parameters include active power P, reactive power Q, head h, and flow rate q. The equipment operating status includes vibration v and swing s. Step S22: Based on the long short-term memory neural network, learn and train the historical operating data of the equipment to obtain the pre-vibration repair state model E of the equipment. v = f(P,Q,h,q) and the slew rate pre-correction state model E s =f(P,Q,h,q), where the pre-maintenance state model reflects the mapping relationship between the equipment's operating status and working parameters before this maintenance.

4. The method for evaluating the maintenance quality of power plant equipment according to claim 1, characterized in that, The operating status weight p mentioned in step S5 is determined using the three-scale method, and the equipment maintenance score DT is calculated using the following formula: Where, p i Let D be the weight of the i-th running state. i Let m be the maintenance score for the i-th operating state, and m be the number of operating states.

5. The method for evaluating the maintenance quality of power plant equipment according to claim 4, characterized in that, The three-scale method for determining state weights includes: Step S51: Construct a three-scale judgment matrix for each running state based on the relative importance of each running state; Step S52: Calculate the weight of each operating state to obtain the feature vector of the decision weight of each level of indicators; Step S53: Perform a consistency check on the weight feature vectors of each operating state; Step S54: If the consistency check meets the standard, determine the weight of each operating state; if it does not meet the standard, adjust the judgment matrix and repeat steps S51 to S53.

Images