A real-time comprehensive evaluation method and system for the health status of a hydropower unit

Abstract

The present invention provides a real-time comprehensive evaluation method and system for the health state of a hydropower unit, including: determining a trained health model; acquiring in real time the active power, water head of the hydropower unit and the real-time state monitoring quantities of each key measuring point, inputting the active power and the water head into the trained health model to obtain the health state monitoring quantities of each key measuring point; respectively constructing Gaussian cloud models for the health state monitoring quantities and the real-time state monitoring quantities of each key measuring point, calculating in real time the cloud similarity degrees of the two Gaussian cloud models of each key measuring point, and determining the health indexes of each key measuring point based on the cloud similarity degrees; wherein, the higher the cloud similarity degree of a certain key measuring point, the smaller its health index value; and fusing the real-time health indexes of each key measuring point based on the weights of each key measuring point to determine the real-time comprehensive health state of the hydropower unit. The present invention can comprehensively evaluate the health state of the unit.

Description

Technical Field

[0001] The present invention belongs to the field of performance monitoring of hydro-generating units, and more specifically, relates to a method and system for real-time comprehensive assessment of the health status of hydro-generating units. Background Art

[0002] As a key device for water energy conversion, hydro-generating units undertake important tasks such as power generation, peak regulation, and frequency modulation in the power system. However, due to the complex and changeable working conditions and harsh working environment of hydro-generating units, with the accumulation of operation time, the performance of each component of the unit gradually deteriorates and even evolves into a fault, affecting the safe, stable, and efficient operation of the unit. In recent years, the condition-based maintenance strategy of hydro-generating units has received increasing attention, avoiding the problems of over-maintenance or untimely maintenance easily caused by traditional planned maintenance and post-fact maintenance, reducing the maintenance cost, and improving the comprehensive benefit of the power station. Therefore, accurately evaluating the real-time health status of the unit is of great significance for effectively carrying out condition-based maintenance and then reducing the failure rate of hydro-generating units.

[0003] Existing research does not consider the influence of working conditions to clean the on-site measured signals of hydro-generating units. At the same time, the health status of the unit is affected by multiple factors, and multi-source evaluation indicators should be considered for a comprehensive evaluation of the health status of the unit during state evaluation. In addition, the real-time change of monitoring signals has uncertainty. When constructing the health indicators of hydro-generating units, the uncertain information should be fully considered, and the essential characteristics and dynamic change characteristics of each indicator should be taken into account during the fusion of multi-source indicators. In order to accurately evaluate the operation status of the unit and provide technical support for condition-based maintenance, it is urgent to carry out real-time comprehensive evaluation of the health status of hydro-generating units. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a method and system for real-time comprehensive assessment of the health status of hydro-generating units, aiming to solve the problems such as insufficient data preprocessing during the evaluation of the health status of existing hydro-generating units, the health model only considering a single measuring point, the construction of health indicators not considering the uncertainty of signal changes, and the inability to comprehensively evaluate the overall status of the unit in real time.

[0005] To achieve the above purpose, in the first aspect, the present invention provides a method for real-time comprehensive assessment of the health status of hydro-generating units, including the following steps:

[0006] Determine a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydro-generating unit and the state monitoring quantities of the unit, and the state monitoring quantities are the vibration quantities or pressure pulsation quantities of each key measuring point under the healthy state of the unit;

[0007] Obtain the real-time active power, water head and real-time state monitoring quantities of each key measuring point of the hydro-generating unit in real time, input the active power and water head into the trained health model, and obtain the health state monitoring quantities of each key measuring point;

[0008] Build Gaussian cloud models for the health status monitoring quantities and real-time status monitoring quantities of each key measurement point respectively, calculate the cloud similarity of the two Gaussian cloud models of each key measurement point in real time, and determine the health index of each key measurement point based on the cloud similarity; among them, the higher the cloud similarity of a certain key measurement point, the smaller its health index value.

[0009] Fuse the real-time health indexes of each key measurement point based on the weights of each key measurement point to determine the real-time comprehensive health status of the hydropower unit; among them, the smaller the value of the real-time comprehensive health status, the better the health status of the hydropower unit.

[0010] In a possible example, the training process of the health model is as follows:

[0011] Obtain the historical operation data of the hydropower unit in the normal state, clean the abnormal points in the data using the multi-scale data cleaning method, and construct an effective health data set; the operation data includes: active power, water head, and the status monitoring quantity of the monitoring point.

[0012] Adopt the comprehensive correlation analysis method to screen the key measurement points most relevant to the operation status of the hydropower unit from the massive monitoring point data as the multi-source health assessment indexes of the unit.

[0013] Use the active power and water head of the hydropower unit in the effective health data set as the working condition parameters to input into the health model, and use the status monitoring quantities of each key measurement point of the hydropower unit as the output labels to train the health models corresponding to each key measurement point.

[0014] In a possible example, the specific steps of using the multi-scale data cleaning method to clean the abnormal points in the data and construct an effective health data set are as follows:

[0015] Input the historical operation data of the hydropower unit in the normal state into the density-based spatial noise clustering algorithm model to initially filter the obvious outliers and obtain relatively dense data samples.

[0016] Input the working condition parameters of the densely filtered data samples after the initial filtering into the Gaussian mixture model to divide them into multiple working condition intervals, use the obtained Gaussian distribution probability density function as the working condition probability of each working condition interval, and use half of the negative logarithmic likelihood probability of the mean of the working condition probability density in each working condition interval as the cleaning density threshold of the vibration signal under this working condition interval; among them, the working condition parameters of the densely filtered data samples include: effective power and water head.

[0017] According to the cleaning density threshold of each working condition interval, use the density-based spatial noise clustering algorithm to clean the densely filtered data samples for each working condition interval respectively, eliminate the dense outliers, and obtain an effective health data set.

[0018] In a possible example, the comprehensive correlation analysis method is adopted to screen the key measuring points most relevant to the operating state of the hydropower unit from the massive monitoring point data, specifically as follows:

[0019] Calculate the Pearson correlation coefficient, the maximum information coefficient, and the grey relational degree between the operating condition parameters of the hydropower unit and the state monitoring quantities at the massive monitoring points respectively, and take the average of the Pearson correlation coefficient, the maximum information coefficient, and the grey relational degree of each monitoring point to obtain the comprehensive correlation degree of the monitoring point;

[0020] Take multiple monitoring points with a comprehensive correlation degree greater than the correlation threshold as the key measuring points.

[0021] In a possible example, Gaussian cloud models of the health state monitoring quantities and the real-time state monitoring quantities of each key measuring point are constructed respectively, the cloud similarity degree of the two Gaussian cloud models of each key measuring point is calculated in real time, and the health index of each key measuring point is determined based on the cloud similarity degree, specifically as follows:

[0022] Input the health state monitoring quantity and the real-time state monitoring quantity of each key measuring point into the Gaussian cloud model respectively to obtain the two Gaussian clouds corresponding to each key measuring point, and determine the three numerical features of the Gaussian cloud: expectation, entropy, and hyperentropy;

[0023] Map the three numerical features of the two Gaussian clouds to the three-dimensional space as three-dimensional coordinates respectively. The two Gaussian clouds mapped to the three-dimensional space can be regarded as two points, and the distance between the two points is calculated using the Euclidean distance;

[0024] Use the distance D of the two points obtained by mapping the Gaussian cloud to the three-dimensional space IJ to represent the similarity degree between the two Gaussian clouds. The greater the distance between the two points, the smaller the similarity degree of the distribution of the two Gaussian clouds; on the contrary, the closer or even the coincidence of the two points, the greater the similarity degree of the distribution of the two Gaussian clouds;

[0025] Determine the health index HI of each key measuring point based on the cloud similarity degree, specifically as follows:

[0026] In a possible example, based on the weights of each key measuring point, fuse the real-time health indexes of each key measuring point to determine the real-time comprehensive health state of the hydropower unit, specifically as follows:

[0027] Calculate the real-time comprehensive health state RCHI of the hydropower unit through the following formula:

[0028]

[0029] where, HI i represents the health index of the i-th key measuring point, represents the weight of the i-th key measuring point, and m represents the total number of key measuring points.

[0030] In a possible example, the weight of the i-th key measurement point is determined through the following steps:

[0031] The self-adjusting analytic hierarchy process is used to determine the prior weight W of each key measurement point pr ;

[0032] The information entropy is used to calculate the dynamic weight W of each key measurement point var , and it is calculated through the following formula:

[0033]

[0034]

[0035] where e i is the information entropy of the i-th evaluation index, N is the number of samples, HI ij is the health index of the i-th key measurement point at the j-th moment, is the dynamic weight of the i-th key measurement point,

[0036] The weights of each key measurement point are calculated through the following formula:

[0037]

[0038] In a second aspect, the present invention provides a real-time comprehensive evaluation system for the health state of a hydropower unit, including:

[0039] A health model determination unit, configured to determine a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydropower unit and the state monitoring quantities, and the state monitoring quantities are the vibration quantities or pressure pulsation quantities of each key measurement point in the healthy state of the unit;

[0040] A health state monitoring unit, configured to obtain the active power, water head and the real-time state monitoring quantities of each key measurement point of the hydropower unit in real time, input the active power and water head into the trained health model, and obtain the health state monitoring quantities of each key measurement point;

[0041] A health index acquisition unit, configured to respectively construct Gaussian cloud models of the health state monitoring quantities and the real-time state monitoring quantities of each key measurement point, and calculate the cloud similarity of the two Gaussian cloud models of each key measurement point in real time, and determine the health index of each key measurement point based on the cloud similarity; among them, the higher the cloud similarity of a certain key measurement point, the smaller its health index value;

[0042] A health status determination unit, which is used to determine the real-time comprehensive health status of a hydropower unit by fusing the real-time health indicators of each key measurement point based on the weights of each key measurement point; wherein, the smaller the value of the real-time comprehensive health status, the better the health status of the hydropower unit.

[0043] In a possible example, the system further includes:

[0044] A health model training unit, which is used to obtain the historical operation data of the hydropower unit in the normal state, clean the abnormal points in the data by using a multi-scale data cleaning method, and construct an effective health data set; the operation data includes: active power, water head, and the status monitoring quantities of the monitoring points; use a comprehensive correlation analysis method to screen the key measurement points most relevant to the operation status of the hydropower unit from the massive monitoring point data as the multi-source health assessment indicators of the unit; and use the active power and water head of the hydropower unit in the effective health data set as the working condition parameters to input into the health model, and use the status monitoring quantities of each key measurement point of the hydropower unit as the output labels to train the health model corresponding to each key measurement point.

[0045] In a possible example, the health model training unit uses a multi-scale data cleaning method to clean the abnormal points in the data and construct an effective health data set. Specifically: input the historical operation data of the hydropower unit in the normal state into a density-based spatial noise clustering algorithm model to initially filter the obvious outliers to obtain relatively dense data samples; input the working condition parameters of the dense data samples after the initial filtering into a Gaussian mixture model to divide them into multiple working condition intervals, use the obtained Gaussian distribution probability density function as the working condition probability of each working condition interval, and use half of the negative logarithmic likelihood probability of the mean of the working condition probability density in each working condition interval as the cleaning density threshold of the vibration signal under this working condition interval; wherein, the working condition parameters of the dense data samples include: effective power and water head; and according to the cleaning density threshold of each working condition interval, use the density-based spatial noise clustering algorithm to clean the dense data samples after the initial filtering for each working condition interval respectively, and eliminate the dense outliers to obtain an effective health data set.

[0046] In a third aspect, the present invention provides an electronic device, including: a memory and a processor;

[0047] The memory is used to store a computer program;

[0048] The processor is used to implement the method provided in the first aspect above when executing the computer program.

[0049] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method provided in the first aspect above is implemented.

[0050] Generally speaking, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:

[0051] The real-time comprehensive evaluation method and system for the health state of a hydropower unit provided by the present invention propose a multi-scale data cleaning method. Considering the operating conditions and changing effects of the unit, based on the probability distribution, time series characteristics, and density distribution characteristics of the measured data of the unit, density thresholds are determined in the joint distribution of head - active power - monitored quantity for different working condition intervals to clean abnormal data, laying a data foundation for the evaluation of the unit's health state.

[0052] The real-time comprehensive evaluation method and system for the health state of a hydropower unit provided by the present invention use a comprehensive correlation analysis method to screen key measurement points that are most closely related to the operating state of the unit from a large amount of multi-source heterogeneous monitoring data, construct multi-source health evaluation indicators, and comprehensively consider the impact of different monitoring positions on the overall unit.

[0053] The real-time comprehensive evaluation method and system for the health state of a hydropower unit provided by the present invention propose a health index calculation method, which fully considers the changes in the health state monitoring signals of the unit and avoids the influence of hyperparameters of conventional methods on the model results. It describes the changes in the monitored quantity of the unit and its uncertainty information from both qualitative and quantitative aspects, and more comprehensively depicts the changes in the actual operating state of the unit.

[0054] The real-time comprehensive evaluation method and system for the health state of a hydropower unit provided by the present invention construct a real-time comprehensive health state of the unit, which integrates the self-adjusting analytic hierarchy process and information entropy, and includes the prior weights and dynamic weights of each index. It comprehensively considers the essential characteristics and dynamic change characteristics of each monitoring index, reflects the real-time changes and inherent characteristics of the unit's health state, and the obtained results are more comprehensive and integrated. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flowchart of a real-time comprehensive evaluation method for the health state of a hydropower unit provided by an embodiment of the present invention;

[0056] Figure 2 It is a flowchart of another real-time comprehensive evaluation method for the health state of a hydropower unit provided by an embodiment of the present invention;

[0057] Figure 3 It is a distribution diagram of active power - head - vibration of on-site measured data of a hydropower unit in an embodiment of the present invention, where (a) is the original data distribution, (b) is the data after primary filtering, and (c) is the data after multi-scale cleaning.

[0058] Figure 4 It is a schematic diagram of the generation process of the Gaussian cloud model in an embodiment of the present invention;

[0059] Figure 5These are the health index curve graphs of key measurement points in the embodiments of the present invention. Among them, (a) is the HI of the Z - direction vibration of the lower bracket LZ , (b) is the HI of the Y - direction vibration of the upper machine frame UY , (c) is the HI of the Y - direction vibration of the top cover TY , (d) is the HI of the pressure pulsation of the top cover TP , (e) is the HI of the pressure pulsation of the draft tube CP , (f) is the HI of the pressure pulsation of the draft elbow EP , (g) is the HI of the pressure pulsation at the outlet of the guide vane GP ;

[0060] Figure 6 This is the real - time comprehensive health status RCHI curve graph in the embodiments of the present invention;

[0061] Figure 7 This is the architecture diagram of the real - time comprehensive health assessment system for the hydropower unit provided by the embodiments of the present invention. Detailed implementation manners

[0062] To make the objectives, 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 with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way restricts the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0063] In view of the above - mentioned defects or improvement requirements of the prior art, the present invention provides a real - time comprehensive health assessment method and system for hydropower units, which can associate multi - scale cleaning of measured data under working conditions, screen key measurement points from a large amount of multi - source heterogeneous data, quantitatively and qualitatively analyze the state changes and uncertainties of hydropower units, and comprehensively and real - time evaluate the health status of the units.

[0064] Figure 1 This is the flowchart of a real - time comprehensive health assessment method for hydropower units provided by the embodiments of the present invention; as Figure 1 shown, it includes the following steps:

[0065] S101, determine a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydropower unit and the state monitoring quantities, and the state monitoring quantities are the vibration quantities or pressure pulsation quantities of each key measurement point under the healthy state of the unit;

[0066] S102. Obtain the active power, water head, and real-time status monitoring quantities of key measurement points of the hydropower unit in real time. Input the active power and water head into the trained health model to obtain the health status monitoring quantities of each key measurement point.

[0067] S103. Construct Gaussian cloud models for the health status monitoring quantities and real-time status monitoring quantities of each key measurement point respectively, calculate the cloud similarity of the two Gaussian cloud models of each key measurement point in real time, and determine the health index of each key measurement point based on the cloud similarity. Among them, the higher the cloud similarity of a certain key measurement point, the smaller its health index value.

[0068] S104. Determine the real-time comprehensive health status of the hydropower unit by fusing the real-time health indexes of each key measurement point based on the weights of each key measurement point. Among them, the smaller the value of the real-time comprehensive health status, the better the health status of the hydropower unit.

[0069] In a specific embodiment, the present invention provides another real-time comprehensive evaluation method for the health status of a hydropower unit, as Figure 2 shown, including the following steps:

[0070] S1: Obtain the monitoring data of the hydropower unit under normal operating conditions, couple the abnormal points in the monitoring data by a multi-scale data cleaning method combined with the working condition conditions, and construct an effective health sample data set.

[0071] S2: Adopt a comprehensive correlation analysis method to screen the key measurement points most relevant to the unit operating status from the massive monitoring data as the multi-source health evaluation indexes of the unit.

[0072] S3: Use the historical health operation condition parameters of the unit as the input and the monitoring quantity under the corresponding working condition as the output label to train the health model.

[0073] S4: Input the real-time collected operation data of the unit into the trained health model to calculate the corresponding health monitoring quantity.

[0074] S5: Construct Gaussian cloud models for the health monitoring quantity and real-time monitoring quantity respectively, calculate the cloud similarity according to the Euclidean distance, and obtain the health index of each measurement point.

[0075] S6: Determine the real-time comprehensive health status of the hydropower unit by fusing the real-time health indexes of each key measurement point based on the weights of each key measurement point

[0076] As a further preference, the specific implementation manner of the multi-scale data cleaning model in step S1 is:

[0077] S101: Input the original sample data (P-H-V) into the density-based spatial noise clustering algorithm model for primary filtering of the obvious outliers, and obtain relatively dense data samples.

[0078] S102: Input the operating condition values (P, H) of the dense data samples after primary filtering into the Gaussian mixture model, divide them into k operating condition intervals, use the obtained Gaussian distribution probability density function as the operating condition probability of each interval, and take half of the negative log-likelihood probability NLL of the mean of the operating condition probability density within each operating condition interval as the cleaning density threshold ε of the vibration signal under this operating condition:

[0079]

[0080] Among them, NLL reflects the likelihood of observing the data on the premise of taking the probability distribution of the operating condition as the prior, and k is the number of divided operating condition intervals;

[0081] S103: Take ε as the cleaning threshold for each operating condition area, and use the density-based spatial noise clustering algorithm to clean the monitoring signals of the unit interval by interval, remove the dense abnormal points, and obtain an effective health data set.

[0082] Preferably, the multi-scale data cleaning model is based on the probability distribution, time series characteristics and density distribution characteristics of the data, cleans the abnormal data in intervals under the joint distribution of water head - active power - monitored quantity, can effectively identify the isolated outliers and dense abnormal points in the measured data, improves the data quality, and lays a data foundation for the health state assessment of the unit.

[0083] As a further preference, the comprehensive correlation analysis method is used to screen the monitoring parameters most relevant to the unit operation from the massive multi-source monitoring data to comprehensively characterize the unit operation state.

[0084] The comprehensive correlation analysis method calculates the Pearson correlation coefficient, the maximum information coefficient and the grey correlation degree between the operating condition parameters of the hydropower unit and the monitored quantity respectively. The obtained comprehensive correlation degree R(V i ) can be calculated by the following formula:

[0085]

[0086]

[0087] Among them, MIC(V i ), PCC(V i ), GCD(V i ) are the Pearson correlation coefficient, the maximum information coefficient and the grey correlation degree of (P, H) and V i respectively, R(V i ) is the comprehensive correlation degree of (P, H) and V i , N is the number of state monitoring quantities, thr is the correlation threshold, and m monitoring quantities with R(V i ) > thr are selected as the key monitoring quantities.

[0088] As a further preference, in step S5, a Gaussian cloud model is introduced to construct a health index based on cloud similarity, including the following steps:

[0089] S501: Input the real-time collected status monitoring quantities and the health status monitoring quantities into the Gaussian cloud model respectively to obtain the real-time Gaussian cloud C(Ex, En, He) and the health Gaussian cloud where Ex, En, and He are the three numerical characteristics of the Gaussian cloud: expectation, entropy, and hyperentropy;

[0090] S502: Map the Gaussian cloud into a three-dimensional space, where the expectation Ex, entropy En, and hyperentropy He are used as the x-axis, y-axis, and z-axis in the three-dimensional coordinate system respectively. The real-time Gaussian cloud C(Ex, En, He) and the health Gaussian cloud mapped into the three-dimensional space can be regarded as two points I(Ex, En, He) and The distance between the two points distance(I|J) can be calculated using the Euclidean distance:

[0091]

[0092] S503: Use the distance between the two points where the Gaussian cloud is mapped into the three-dimensional space to represent the similarity between the two Gaussian clouds. The greater the distance between the two points, the smaller the similarity of the two cloud distributions; conversely, the closer or even coincident the two points are, the greater the similarity of the cloud distributions. Denote the cloud similarity as D IJ :

[0093] D IJ = distance(I|J)

[0094] S504: Accordingly, construct a health index HI, which can be calculated by the following formula:

[0095]

[0096] where HI is the health index of the unit at the current moment, and D IJ is the cloud similarity between the real-time Gaussian cloud and the health Gaussian cloud. Calculate the health indexes of m key measuring points respectively to obtain the multi-source health index of the hydropower unit;

[0097] Preferably, the Gaussian cloud describes the randomness and ambiguity of a concept from a quantitative perspective. The expectation is the expected value of the distribution of cloud droplets in the universe of discourse space, which describes the most representative and typical samples and corresponds to the width of the cloud droplets. The entropy reflects the uncertainty of the distribution of cloud droplets, corresponding to the dispersion degree of cloud droplets, which is jointly determined by the degree of discreteness and fuzziness. The hyper-entropy is the uncertainty of the entropy, corresponding to the thickness of the cloud droplets. The uncertainty of the change of the monitoring signal is quantified by the Gaussian cloud. The distribution difference between the real-time Gaussian cloud of the real-time online monitoring data and the healthy Gaussian cloud under the corresponding working conditions is used as the change of the health state of the unit. Based on this, a health index is constructed to describe the change of the monitored quantity of the unit and its certainty information from both qualitative and quantitative aspects, and at the same time, the influence of model hyperparameters on the calculation is avoided.

[0098] Further preferably, in step S6, the real-time comprehensive health state of the unit is obtained by calculating the hybrid weights to fuse the health indexes of each measuring point. Its characteristics are as follows: the hybrid weights include prior weights and dynamic weights, comprehensively considering the inherent attributes and dynamic changes of each monitoring index, and comprehensively representing the real-time change of the health state of the unit.

[0099] The calculation steps of the hybrid weights are as follows:

[0100] S601: Combining expert experience and the unit maintenance records, the prior weight Wi of each index is calculated by using the self-adjusting analytic hierarchy process. Wi reflects the inherent attributes of different indexes. pr Wi pr reflects the inherent attributes of different indexes;

[0101] S602: The dynamic weight Wi of each index is calculated by using information entropy. Wi reflects the real-time change characteristics of each index. Wi can be calculated by the following formula: var Wi var reflects the real-time change characteristics of each index. Wi var can be calculated by the following formula:

[0102]

[0103]

[0104] where, ei i is the information entropy of the i-th evaluation index, N is the number of samples, m is the number of evaluation indexes, HIij ij is the HI of the i-th evaluation index at the j-th moment. HI may be zero. Let 0ln0 = 0. Wij is the dynamic weight of the i-th evaluation index.

[0105] S603: The hybrid weight Wi of each index can be calculated by the following formula: mix can be calculated by the following formula:

[0106]

[0107] S604: The real-time comprehensive health status RCHI of the hydropower unit is calculated by the following formula:

[0108]

[0109] Further preferably, the self-adjusting analytic hierarchy process in step S601 includes the following steps:

[0110] (1) For m evaluation indicators I = (I1, I2, ..., I m ), based on the industry experts’ experience, unit failure records and maintenance records, determine the comparison matrix C = (c ab ) m*m , where c ab Defined as:

[0111]

[0112] (2) Divide the indicator set:

[0113]

[0114]

[0115] Where × represents the Cartesian product, D, H, M, L are process matrices, and DL, DM, DH are process sets;

[0116] (3) If DL a DM a and DH i are all empty sets, go to step (4), otherwise let:

[0117]

[0118] If a=m-1, return to step (2); otherwise, set a=a+1 and return to step (1);

[0119] (4) The other elements of the comparison matrix C can be determined by the following formula:

[0120]

[0121] (5) Find the optimal transfer matrix U of C = (u ab ) m×m , where u ab Defined as:

[0122]

[0123] (6) Find U = (u ab ) m×m The maximum eigenvalue of , normalize the eigenvector corresponding to the maximum eigenvalue, and get the index (I1,I2,…,Im ) prior weight W pr =(w pr1 , w pr2 ,..., w prm ).

[0124] Preferably, the self - adjusting hierarchy method greatly simplifies the calculation steps of the analytic hierarchy process, has self - adaptability, and can objectively reflect the relevant expert experience in the assessment of the unit health status, as well as the unit's historical fault records and maintenance records.

[0125] The real - time comprehensive assessment method for the health status of hydropower units provided by the present invention aims to correlate the measured data of multi - scale cleaning under working conditions, screen key measuring points from a large amount of multi - source heterogeneous data, quantitatively and qualitatively analyze the state changes and their uncertainties of hydropower units, and comprehensively and real - time assess the health status of the units.

[0126] The multi - scale data cleaning method used in the present invention is based on the joint distribution of active power - water head - vibration of measured data, and cleans abnormal data in different working condition intervals to provide a data basis for health status assessment. As Figure 3 shown in the result diagram of the multi - scale data cleaning method in the embodiment of the present invention.

[0127] The unit state monitoring signal is not a definite value under specific working conditions, but shows a specific distribution. The Gaussian cloud model used in the present invention can describe the changes and their uncertainty information of the monitoring signal from both qualitative and quantitative aspects. As Figure 4 shown in the process of the Gaussian cloud model describing the monitoring signal in the embodiment of the present invention.

[0128] Calculate the similarity of the Gaussian cloud model for each key measuring point to obtain the health index curves of each measuring point as Figure 5 shown. It can be seen from Figure 5 that the fluctuation amplitudes of the health curves of different measuring points are different, but the overall trend is upward, reflecting the real - time change process of the unit health status. Among them, the upward trend of the health curve indicates that as the unit operation time increases, different parts of the unit show different degrees of aging, that is, the health status decreases as the unit operation time increases.

[0129] Calculate the mixed weights of each measuring point in the embodiment of the present invention, and finally obtain the real - time comprehensive health status curve of the unit as Figure 6 shown. It can be seen that this curve synthesizes the characteristics of each measuring point, shows an overall local fluctuation and a gradually upward trend, which conforms to the actual situation of the unit health status, that is, as the unit operation time increases, the health status of the unit decreases.

[0130] Verification shows that: using the real - time comprehensive assessment method for the health status of hydropower units of the present invention, the change curve of the unit health status can be obtained based on the on - site monitoring data of hydropower units, and the health status of the unit can be comprehensively and real - time assessed.

[0131] Figure 7 This is the architecture diagram of the real-time comprehensive health assessment system for hydropower units provided by the embodiments of the present invention. As Figure 7 shown, it includes:

[0132] A health model determination unit 710, configured to determine a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydropower unit and the unit state monitoring quantities, and the state monitoring quantities are the vibration quantities or pressure pulsation quantities of each key measurement point under the healthy state of the unit;

[0133] A health state monitoring unit 720, configured to obtain the real-time active power, water head of the hydropower unit and the real-time state monitoring quantities of each key measurement point in real time, input the active power and water head into the trained health model, and obtain the health state monitoring quantities of each key measurement point;

[0134] A health index acquisition unit 730, configured to respectively construct Gaussian cloud models of the health state monitoring quantities and real-time state monitoring quantities of each key measurement point, and calculate the cloud similarity of the two Gaussian cloud models of each key measurement point in real time, and determine the health index of each key measurement point based on the cloud similarity; among them, the higher the cloud similarity of a certain key measurement point, the smaller its health index value;

[0135] A health state determination unit 740, configured to fuse the real-time health indexes of each key measurement point based on the weights of each key measurement point to determine the real-time comprehensive health state of the hydropower unit; among them, the smaller the value of the real-time comprehensive health state, the better the health state of the hydropower unit;

[0136] A health model training unit 750, configured to obtain the historical operation data of the hydropower unit in the normal state, clean the abnormal points in the data by using a multi-scale data cleaning method, and construct an effective health data set; the operation data includes: active power, water head and the state monitoring quantities of the monitoring points; use a comprehensive correlation analysis method to screen the key measurement points most relevant to the operation state of the hydropower unit from the massive monitoring point data as the multi-source health assessment indexes of the unit; and use the active power and water head of the hydropower unit in the effective health data set as the working condition parameters to input into the health model, and use the state monitoring quantities of each key measurement point of the hydropower unit as the output labels to train the health models corresponding to each key measurement point.

[0137] It can be understood that the detailed function implementation of each of the above units can be referred to the introduction in the foregoing method embodiments, and will not be elaborated here.

[0138] The present invention discloses a real-time comprehensive evaluation method and system for the health status of a hydropower unit, which relates to the technical field of hydropower unit status evaluation. The method is based on the historical monitoring data of the normal operation state of the hydropower unit. First, a multi-scale data cleaning method is used to process the measured data considering the working conditions, and the abnormal points are removed to provide a data basis for status evaluation. Then, a comprehensive correlation analysis method is used to screen key measurement points reflecting the unit operation state from the massive multi-source monitoring data, and a multi-source evaluation index is constructed. Further, a health index is calculated based on the Euclidean distance-Gaussian cloud model. Finally, a hybrid weight calculation method is proposed to fuse the multi-source evaluation indexes to obtain the real-time comprehensive health status of the unit. The present invention can effectively clean abnormal data, quantify the uncertain information of the signal, and the constructed real-time comprehensive health status can comprehensively characterize the health status change of the hydropower unit, providing technical support for condition-based maintenance.

[0139] In addition, an embodiment of the present invention provides an electronic device, which includes: a memory and a processor;

[0140] The memory is used to store a computer program;

[0141] The processor is used to implement the method in the above embodiment when executing the computer program.

[0142] Furthermore, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method in the above embodiment is implemented.

[0143] Based on the method in the above embodiment, an embodiment of the present invention provides a computer program product, and when the computer program product runs on a processor, the processor is caused to execute the method in the above embodiment.

[0144] Based on the method in the above embodiment, an embodiment of the present invention further provides a chip, which includes one or more processors and an interface circuit. Optionally, the chip may further include a bus. Among them:

[0145] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit in the hardware of the processor or the instructions in the form of software. The above processor may be a general-purpose processor, a digital communicator (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The various methods and steps disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0146] The interface circuit can be used for sending or receiving data, instructions or information. The processor can utilize the data, instructions or other information received by the interface circuit for processing, and can send out the processed information through the interface circuit.

[0147] Optionally, the chip further includes a memory, which can include a read-only memory and a random access memory, and provides operation instructions and data to the processor. A part of the memory can also include a non-volatile random access memory (NVRAM).

[0148] Optionally, the memory stores executable software modules or data structures, and the processor can execute corresponding operations by calling the operation instructions stored in the memory (the operation instructions can be stored in the operating system).

[0149] Optionally, the interface circuit can be used to output the execution result of the processor.

[0150] It should be noted that the functions corresponding to the processor and the interface circuit can be implemented through hardware design, software design, or a combination of hardware and software, and there is no limitation here.

[0151] It should be understood that each step of the above method embodiments can be completed by the logic circuit in hardware form or the instructions in software form in the processor.

[0152] It can be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application. In addition, in some possible implementation manners, the steps in the above embodiments can be selectively executed according to the actual situation, can be partially executed, or can be fully executed, and there is no limitation here.

[0153] It can be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0154] The method steps in the embodiments of this application can be implemented in a hardware manner or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a random access memory (RAM), flash memory, read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), register, hard disk, removable hard disk, CD-ROM, or any other form of storage medium well-known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0155] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server, data center, etc. that includes one or more integrated available media. The available medium can be a magnetic medium (for example, floppy disk, hard disk, magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, solid state disk (SSD)), etc.

[0156] Those skilled in the art can easily understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time comprehensive evaluation method for the health status of a hydropower unit, characterized in that, It includes the following steps: Determine a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydropower unit and the unit state monitoring quantity, and the state monitoring quantity is the vibration quantity or pressure pulsation quantity of each key measuring point in the healthy state of the unit; the training process of the health model is as follows: Obtain the historical operation data of the hydropower unit in the normal state, clean the abnormal points in the data by using a multi-scale data cleaning method, and construct an effective health data set; the operation data includes: active power, water head and the state monitoring quantity of the monitoring points; Adopt a comprehensive correlation analysis method to screen the key measuring points most relevant to the operation state of the hydropower unit from the massive monitoring point data as the multi-source health evaluation indexes of the unit; Use the active power and water head of the hydropower unit in the effective health data set as the operating condition parameters to input into the health model, and use the state monitoring quantity of each key measuring point of the hydropower unit as the output label to train the health model corresponding to each key measuring point; Obtain the active power, water head and the real-time state monitoring quantity of each key measuring point of the hydropower unit in real time, input the active power and water head into the trained health model, and obtain the health state monitoring quantity of each key measuring point; Construct Gaussian cloud models for the health state monitoring quantity and the real-time state monitoring quantity of each key measuring point respectively, calculate the cloud similarity of the two Gaussian cloud models of each key measuring point in real time, and determine the health index of each key measuring point based on the cloud similarity; specifically: Input the health state monitoring quantity and the real-time state monitoring quantity of each key measuring point into the Gaussian cloud model respectively to obtain two Gaussian clouds corresponding to each key measuring point, and determine three numerical characteristics of the Gaussian cloud: expectation, entropy and hyper-entropy; Map the three numerical characteristics of the two Gaussian clouds to the three-dimensional space as three-dimensional coordinates respectively. The two Gaussian clouds mapped to the three-dimensional space can be regarded as two points, and the distance between the two points is calculated by using the Euclidean distance; Distance between two points mapped by Gaussian cloud into three-dimensional space Represents the similarity between two Gaussian clouds. The greater the distance between the two points, the smaller the similarity of the distributions of the two Gaussian clouds; conversely, the closer or even the coincidence of the two points, the greater the similarity of the distributions of the two Gaussian clouds; Determine the health indicators of each key measurement point based on cloud similarity , specifically as follows: ; Among them, the higher the cloud similarity of a certain key measuring point, the smaller its health index value; Fuse the real-time health indexes of each key measuring point based on the weights of each key measuring point to determine the real-time comprehensive health state of the hydropower unit; among them, the smaller the value of the real-time comprehensive health state, the better the health state of the hydropower unit.

2. The method according to claim 1, wherein The method of using a multi-scale data cleaning method to clean the abnormal points in the data and construct an effective health data set is specifically: Input the historical operation data of the hydropower unit in the normal state into the density-based spatial noise clustering algorithm model to initially filter the obvious outliers and obtain relatively dense data samples; Input the operating condition parameters of the densely filtered data samples into the Gaussian mixture model to divide them into multiple operating condition intervals, use the obtained Gaussian distribution probability density function as the operating condition probability of each operating condition interval, and use half of the negative logarithmic likelihood probability of the mean of the operating condition probability density in each operating condition interval as the cleaning density threshold of the vibration signal under this operating condition interval; among them, the operating condition parameters of the densely filtered data samples include: effective power and water head; According to the cleaning density threshold of each operating condition interval, use the density-based spatial noise clustering algorithm to clean the densely filtered data samples separately for each operating condition interval, eliminate the dense outliers, and obtain an effective health data set.

3. The method according to claim 1, wherein The method of using comprehensive correlation analysis to screen key measurement points most relevant to the operating state of a hydropower unit from a large amount of monitoring point data is as follows: Calculate the Pearson correlation coefficient, the maximum information coefficient, and the grey relational degree between the operating condition parameters of the hydropower unit and the state monitoring quantities at a large number of monitoring points respectively, and take the average value of the Pearson correlation coefficient, the maximum information coefficient, and the grey relational degree of each monitoring point to obtain the comprehensive correlation degree of this monitoring point; Take multiple monitoring points with a comprehensive correlation degree greater than the correlation threshold as key measurement points.

4. The method according to any one of claims 1 to 3, characterized in that, Based on the weights of each key measurement point, fuse the real-time health indicators of each key measurement point to determine the real-time comprehensive health state of the hydropower unit, specifically: The real-time comprehensive health status of the hydropower unit is calculated by the following formula :[[]]END]] Among them, represents the health index of the i th key measurement point, represents the weight of the i th key measurement point, m represents the total number of key measurement points.

5. The method according to claim 4, characterized in that, The i weights of the key measurement points are determined through the following steps: The self - adjusting analytic hierarchy process is adopted to determine the prior weights of each key measuring point ; Calculate the dynamic weights of each key measurement point using information entropy , which is calculated by the following formula: Among them, is the information entropy of the th evaluation index, is the number of samples, is the health index of the th key measurement point at the th moment, is the dynamic weight of the th key measurement point, ; The weights of each key measurement point are calculated by the following formula: 。 6. A real-time comprehensive evaluation system for the health status of a hydropower unit, characterized in that, Including: A health model determination unit for determining a trained health model; the health model is used to fit the mapping relationship between the active power, water head of the hydropower unit and the state monitoring quantities; the state monitoring quantities are the vibration quantities or pressure pulsation quantities of each key measurement point in the healthy state of the unit; the training process of the health model is as follows: Obtain the historical operation data of the hydropower unit in the normal state, use the multi-scale data cleaning method to clean the abnormal points in the data, and construct an effective health data set; the operation data includes: active power, water head, and the state monitoring quantities of the monitoring points; Use the comprehensive correlation analysis method to screen key measurement points most relevant to the operating state of the hydropower unit from a large amount of monitoring point data as the multi-source health assessment index of the unit; Take the active power and water head of the hydropower unit in the effective health data set as the operating condition parameters and input them into the health model, and take the state monitoring quantities of each key measurement point of the hydropower unit as the output labels to train the health model corresponding to each key measurement point; A health state monitoring unit for obtaining the active power, water head and the real-time state monitoring quantities of each key measurement point of the hydropower unit in real time, inputting the active power and water head into the trained health model, and obtaining the health state monitoring quantities of each key measurement point; A health indicator acquisition unit for respectively constructing Gaussian cloud models of the health state monitoring quantities and the real-time state monitoring quantities of each key measurement point, and calculating the cloud similarity of the two Gaussian cloud models of each key measurement point in real time, and determining the health indicators of each key measurement point based on the cloud similarity; specifically: Input the health state monitoring quantities and the real-time state monitoring quantities of each key measurement point into the Gaussian cloud model respectively, obtain two Gaussian clouds corresponding to each key measurement point, and determine three numerical characteristics of the Gaussian cloud: expectation, entropy, and hyperentropy; Map the three numerical characteristics of the two Gaussian clouds to three-dimensional space respectively as three-dimensional coordinates. The two Gaussian clouds mapped to three-dimensional space can be regarded as two points, and the distance between the two points is calculated using the Euclidean distance; The distance between two points mapped by Gaussian cloud into three-dimensional space represents the similarity between two Gaussian clouds. The greater the distance between the two points, the smaller the similarity of the distributions of the two Gaussian clouds; conversely, the closer or even the coincidence of the two points, the greater the similarity of the distributions of the two Gaussian clouds. Determine the health indicators of each key measurement point based on cloud similarity , specifically as follows: ; Among them, the higher the cloud similarity of a certain key measurement point, the smaller its health indicator value; A health state determination unit for fusing the real-time health indicators of each key measurement point based on the weights of each key measurement point to determine the real-time comprehensive health state of the hydropower unit; among them, the smaller the value of the real-time comprehensive health state, the better the health state of the hydropower unit.

7. The system according to claim 6, wherein It also includes: A health model training unit, which is used to obtain the historical operation data of the hydropower unit in the normal state, clean the abnormal points in the data by using a multi-scale data cleaning method, and construct an effective health data set; the operation data includes: active power, water head, and the monitoring point status monitoring quantity; a comprehensive correlation analysis method is used to screen the key measurement points most relevant to the operation state of the hydropower unit from the massive monitoring point data as the multi-source health evaluation index of the unit; and the active power and water head of the hydropower unit in the effective health data set are used as the working condition parameters to input into the health model, and the status monitoring quantities of each key measurement point of the hydropower unit are used as the output labels to train the health model corresponding to each key measurement point.

8. The system according to claim 6 or 7, characterized in that, The health model training unit uses a multi-scale data cleaning method to clean the abnormal points in the data and construct an effective health data set. Specifically: the historical operation data of the hydropower unit in the normal state is input into the density-based spatial noise clustering algorithm model to initially filter the obvious outliers to obtain relatively dense data samples; the working condition parameters of the dense data samples after the initial filtering are input into the Gaussian mixture model to be divided into multiple working condition intervals, and the obtained Gaussian distribution probability density function is used as the working condition probability of each working condition interval running, and half of the negative logarithmic likelihood probability of the working condition probability density mean in each working condition interval is used as the cleaning density threshold of the vibration signal under this working condition interval; among them, the working condition parameters of the dense data samples include: effective power and water head; and according to the cleaning density threshold of each working condition interval, the density-based spatial noise clustering algorithm is used to clean the dense data samples after the initial filtering for each working condition interval respectively, and the dense outliers are removed to obtain an effective health data set.

Images