Fault Warning Method and System for Hydraulic Generator Sets Based on Digital Twin

Abstract

The present invention discloses a fault warning method and system for a water conservancy generator set based on digital twins, which relates to the technical field of fault analysis of water conservancy generator sets. Specifically, it discloses that state parameters are configured for each virtual component structure in the virtual water conservancy generator set structure, the historical state data is screened for consistency, and the reference historical state data with the highest degree of consistency with the real-time state data is obtained. The virtual dynamic performance of the virtual water conservancy generator set structure corresponding to the reference historical state data is taken as the reference virtual dynamic performance. The real-time sensing data in the real-time state data is compared with the reference virtual dynamic performance, and based on the comparison result, the virtual component structures in the abnormal state on the virtual water conservancy generator set structure are marked. Through the above technical solutions, the present invention realizes the analysis of the real-time state data of the water conservancy generator set, and determines the components with abnormalities based on the analysis results. Compared with manual inspection, it not only has higher efficiency but also higher accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of fault analysis of water conservancy generating units, and particularly to a fault warning method and system for water conservancy generating units based on digital twin. Background Art

[0002] With the rapid development of China's economy, the demand for electricity is increasing. As an important part of clean energy, water conservancy power generation occupies an important position in China's energy structure. However, during the operation of water conservancy generating units, affected by many factors, such as wear, corrosion, loosening and other problems of key components such as water turbines, generators, bearings, transformers, etc., it is easy to cause equipment failures and even accidents. Therefore, it is of great practical significance to conduct fault warning on water conservancy generating units to ensure the safe and stable operation of equipment.

[0003] Traditional fault warning methods mainly rely on manual inspections and experience judgments, which have great limitations. On the one hand, manual inspections cannot grasp the operation status of equipment in real time, and it is difficult to detect potential fault hazards; on the other hand, relying on experience judgments is easily affected by subjective factors, and the warning accuracy is not high. Summary of the Invention

[0004] The purpose of the present invention is to provide a fault warning method and system for water conservancy generating units with high accuracy and capable of automatically judging faults.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A fault warning method for water conservancy generating units based on digital twin, including:

[0007] Obtain the structural information of the water conservancy generating unit, and based on the structural information of the water conservancy generating unit, construct the virtual component structure of the water conservancy generating unit, combine the virtual component structures, and define the motion form of each virtual component structure and the motion range for the motion form to obtain the virtual water conservancy generating unit structure;

[0008] Obtain the historical state data of the water conservancy generating unit, where the historical state data includes historical sensing data and historical load data, and perform independent screening on the data in the historical state data to obtain several sub-historical state data;

[0009] Based on each sub-historical state data, configure the state parameters of each virtual component structure in the virtual water conservancy generating unit structure to obtain the virtual dynamic performance of the virtual water conservancy generating unit structure;

[0010] Obtain the real-time status data of the hydraulic generator set, and based on the real-time load data in the real-time status data, perform consistency screening on the historical status data to obtain the reference historical status data with the highest degree of consistency with the real-time status data, and present the virtual dynamic performance of the virtual hydraulic generator set structure corresponding to the reference historical status data as the reference virtual dynamic performance;

[0011] Compare the real-time sensing data in the real-time status data with the reference virtual dynamic performance, and identify the virtual components with difference parameters greater than or equal to the preset value as abnormal states, and mark the virtual component structures in the virtual hydraulic generator set structure that are in abnormal states.

[0012] In some embodiments disclosed in the present invention, the method for constructing the virtual hydraulic generator set structure includes:

[0013] Construct a three-dimensional coordinate system, and configure different virtual component structures in the three-dimensional coordinate system for combination;

[0014] Based on the movement range of the movement form of the virtual component structure, delimit a movement space in the three-dimensional coordinate system, and delimit a safe movement subspace and several levels of abnormal movement subspaces for the movement space. Based on the status data of the virtual component structure, drive the virtual component structure to move in the corresponding movement subspace.

[0015] In some embodiments disclosed in the present invention, the historical sensing data includes: water turbine temperature data, generator temperature data, bearing temperature data, transformer temperature data, bearing vibration data, rotor vibration data, stator vibration data, water turbine inlet and outlet pressure data, pipeline system pressure data, water turbine flow data, axial displacement data of the rotor, voltage data, current data, lubricating system oil pressure data, lubricating system oil temperature data.

[0016] In some embodiments disclosed in the present invention, the method for performing independence screening on the data in the historical status data includes:

[0017] Based on the analysis of the historical sensing data and the historical status data, determine the historical sensing parameters and historical load parameters at different time nodes in the same time period, and construct several historical sensing parameter curves and historical load parameter curves therefrom. Denote the combination of the historical sensing parameter curve and the historical load parameter curve in the same time period as the historical parameter curve group;

[0018] Performing grid dimension reduction on each historical sensor parameter curve and historical load parameter curve in the historical parameter curve group, the grid dimension reduction method comprising mapping the historical sensor parameter curve or the historical load parameter curve into a grid array template, and recording the dimension reduction grids mapped by the historical sensor parameter curve or the historical load parameter curve in the grid array to obtain a grid array diagram, and recording the combination of the grid array diagrams corresponding to the historical parameter curve group as a grid array diagram group;

[0019] Calculating the average square array graph of the square array graphs of the same type between the square array graph groups, and analyzing the trend labels of the corresponding square array graphs in each square array graph group relative to the average square array graph, the method for determining the trend label includes analyzing the number of first squares in which the dimensionality reduction squares mapped in the square array graph are higher than the dimensionality reduction squares mapped in the average square array graph, and analyzing the number of second squares in which the dimensionality reduction squares mapped in the square array graph are lower than the dimensionality reduction squares mapped in the average square array graph, and identifying the ratio of the first square number to the second square number as the trend label;

[0020] A plurality of square quantity ratio intervals are set for the trend labels of each type of square array graph, and the combination of square quantity ratio intervals corresponding to different types of square array graphs is recorded as a ratio interval classification group;

[0021] Using the ratio interval classification group as the classification basis, different grid array groups are analyzed and classified, and the grid array groups in each category set are completely compared, and repeated grid array groups are eliminated;

[0022] The historical parameter curve group corresponding to the eliminated grid array graph group is identified as the sub-historical state data that has been screened for independence.

[0023] In some embodiments disclosed in the present invention, a method for removing repeated grid array graph groups includes:

[0024] The square array graph groups belonging to the same category set are combined in pairs, and the similarity between the square array graph groups in each pair of combinations is analyzed. If the similarity is greater than or equal to a preset value, one of the square array graph groups is randomly eliminated.

[0025] In some embodiments disclosed in the present invention, the method for calculating the similarity between grid array graph groups includes:

[0026] The same type of square array diagrams between the square array diagram groups are matched one by one, and the dimension reduction squares between the corresponding square array diagrams are analyzed to determine the number of matching squares and the number of non-matching squares;

[0027] Calculate the sum of the number of non - matching squares and the number of non - matching squares to obtain the total number of squares, and determine the similarity degree between the groups of square array diagrams based on the ratio relationship between the number of non - matching squares and the total number of squares, as well as the continuity characteristics of the non - matching squares;

[0028] Among them, the expression for calculating the similarity degree between the groups of square array diagrams is:

[0029]

[0030] Among them, S is the similarity degree between the groups of square array diagrams, K i is the weight coefficient of the i - th type of square array diagram between the groups of square array diagrams, S max is the preset maximum similarity degree, x 1～i is the number of matching squares of the i - th type between the groups of square array diagrams, x 2～i is the number of non - matching squares of the i - th type between the groups of square array diagrams, β[i] is the continuity judgment function of the non - matching squares of the i - th type between the groups of square array diagrams, generate the corresponding continuity correction and adjustment coefficient based on the continuous number of non - matching squares, and n is the number of types of square array diagrams between the groups of square array diagrams;

[0031] Among them, the expression of the continuity judgment function is:

[0032]

[0033] Among them, L is the continuity conversion coefficient, α is the continuous number of non - matching squares, γ is the number of continuous sections of non - matching squares, and b is the continuity adjustment constant.

[0034] In some embodiments disclosed by the present invention, the method for performing consistency screening on historical state data includes:

[0035] Perform hourly analysis on the historical load data and real - time load data in the historical state data, and based on the analysis results, construct a real - time load curve and a historical load curve;

[0036] Establish a time reference axis for the real - time load curve and the historical load curve, and set several time reference sections on the time reference axis, align the real - time load curve, the historical load curve and the time reference axis, and calculate the average value of the real - time load curve and the average value of the historical load curve mapped by each time reference section;

[0037] Compare the average value of the real-time load curve and the average value of the historical load curve successively. If the accumulated difference in average values is greater than or equal to the preset value, replace the historical state data corresponding to the historical load curve until, on the time reference axis scale, the accumulated difference in average values is still less than the preset value. Then, use the corresponding historical state data as the calculation object for the real-time load data to calculate the degree of coincidence.

[0038] In some embodiments disclosed by the present invention, the expression for calculating the degree of coincidence is:

[0039]

[0040] where W is the degree of coincidence, W max is the preset maximum degree of coincidence, R is the difference degree conversion coefficient, Δh g is the cross-sectional area of the g-th intersection interval between the real-time load curve and the historical load curve, G is the total number of intersection intervals between the real-time load curve and the historical load curve, μ is the variance between the real-time load curve and the historical load curve, and c is the variance adjustment constant.

[0041] In some embodiments disclosed by the present invention, there is also disclosed a digital twin-based fault warning system for a water conservancy generator set, including:

[0042] A first module for obtaining the structural information of the water conservancy generator set, constructing the virtual component structure of the water conservancy generator set based on the structural information of the water conservancy generator set, combining the virtual component structures, and defining the motion form of each virtual component structure and the motion range for the motion form to obtain the virtual water conservancy generator set structure;

[0043] A second module for obtaining the historical state data of the water conservancy generator set, where the historical state data includes historical sensing data and historical load data, and performing independence screening on the data in the historical state data to obtain several sub-historical state data;

[0044] A third module for configuring the state parameters of each virtual component structure in the virtual water conservancy generator set structure based on each sub-historical state data to obtain the virtual dynamic performance of the virtual water conservancy generator set structure;

[0045] A fourth module for obtaining the real-time state data of the water conservancy generator set (including real-time sensing data and real-time load data), and performing coincidence screening on the historical state data based on the real-time load data in the real-time state data to obtain the reference historical state data with the highest degree of coincidence with the real-time state data, and using the virtual dynamic performance of the virtual water conservancy generator set structure corresponding to the reference historical state data as the reference virtual dynamic performance;

[0046] The fifth module is used to compare the real-time sensing data in the real-time status data with the reference virtual dynamic performance, identify the virtual components with the difference parameter greater than or equal to the preset value as abnormal states, and mark the virtual component structures in the virtual hydro-generator unit structure that are in abnormal states.

[0047] The present invention discloses a fault warning method and system for a hydro-generator unit based on digital twin, which relates to the technical field of fault analysis of hydro-generator units. Specifically, it discloses configuring state parameters for each virtual component structure in the virtual hydro-generator unit structure, performing consistency screening on historical status data to obtain the reference historical status data with the highest degree of consistency with the real-time status data, and using the virtual dynamic performance of the virtual hydro-generator unit structure corresponding to the reference historical status data as the reference virtual dynamic performance. Comparing the real-time sensing data in the real-time status data with the reference virtual dynamic performance, and marking the virtual component structures in the virtual hydro-generator unit structure that are in abnormal states based on the comparison result. Through the above technical solutions, the present invention realizes the analysis of the real-time status data of the hydro-generator unit, and determines the abnormal components based on the analysis result. Compared with manual inspection, it not only has higher efficiency but also higher accuracy. Description of the Drawings

[0048] Figure 1 This is the fault warning method and system for a hydro-generator unit based on digital twin disclosed in the embodiments of the present invention. Detailed Embodiments

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the 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 the embodiments.

[0050] To achieve the above object, the present invention adopts the following technical solutions:

[0051] The fault warning method for a hydro-generator unit based on digital twin includes:

[0052] Step S100: Obtain the structural information of the hydro-generator unit, and based on the structural information of the hydro-generator unit, construct the virtual component structures of the hydro-generator unit, combine the virtual component structures, and define the motion form of each virtual component structure and the motion range for the motion form to obtain the virtual hydro-generator unit structure.

[0053] In this step, the structural information of the hydraulic generator unit is collected, including the shapes, dimensions, materials, etc. of each component; then, using this information, a virtual structural model of the hydraulic generator unit components is constructed on a computer; these virtual component structures are combined in a three-dimensional coordinate system according to their positions and connection relationships in actual operation; then, the motion forms and motion ranges of each virtual component structure are defined, such as rotation, vibration, etc., so as to form a complete virtual hydraulic generator unit structure.

[0054] In some embodiments disclosed by the present invention, the method for constructing a virtual hydraulic generator unit structure includes:

[0055] Step S101, construct a three-dimensional coordinate system and configure different virtual component structures in the three-dimensional coordinate system for combination.

[0056] The construction of the three-dimensional coordinate system is the basis for creating a virtual environment, which defines the positions and directions in the virtual space; in this coordinate system, each virtual component has a definite coordinate, which corresponds to the physical position of the component in the real world.

[0057] Configuring different virtual component structures in the three-dimensional coordinate system means placing these components in the correct positions according to the actual assembly relationship and spatial layout; in this way, the virtual generator unit can reflect the spatial structure and mutual relationship of the real generator unit.

[0058] Step S102, based on the motion forms and motion ranges of the virtual component structures, delimit a motion space in the three-dimensional coordinate system, and delimit a safe motion subspace and several levels of abnormal motion subspaces for the motion space. Based on the state data of the virtual component structures, drive the virtual component structures to move in the corresponding motion subspaces.

[0059] Based on the motion forms and motion ranges of the virtual component structures, delimit a motion space in the three-dimensional coordinate system; this space defines the area where each component can move during the simulation.

[0060] Within the motion space, further delimit a safe motion subspace and an abnormal motion subspace; the safe motion subspace refers to the motion range that the component should maintain under normal working conditions, while the abnormal motion subspace refers to the motion range of the component in case of possible failures or abnormalities.

[0061] Driving the virtual component structures to move in the corresponding motion subspaces means simulating the actual motion of the components according to the state data (such as sensing data) of the virtual component structures; if the state data indicates that the component motion exceeds the safe motion subspace and enters the abnormal motion subspace, it may indicate the existence of a failure or potential problem.

[0062] In some embodiments disclosed by the present invention, the historical sensing data includes: water turbine temperature data, generator stage temperature data, bearing temperature data, transformer temperature data, bearing vibration data, rotor vibration data, stator vibration data, water turbine inlet and outlet pressure data, pipeline system pressure data, water turbine flow data, axial displacement data of the rotor, voltage data, current data, lubricating system oil pressure data, and lubricating system oil temperature data.

[0063] Step S200: Obtain the historical state data of the hydraulic generating unit. The historical state data includes historical sensing data and historical load data, and perform independence screening on the data in the historical state data to obtain several sub-historical state data.

[0064] In this step, obtain the historical state data from the historical operation records of the hydraulic generating unit, including sensing data (such as temperature, vibration, pressure, etc.) and load data; perform independence screening on these data. The purpose is to remove redundant and highly correlated data and retain the data that can reflect the independent characteristics of the equipment state, thereby obtaining a sub-historical state data set.

[0065] In some embodiments disclosed by the present invention, the method for performing independence screening on the data in the historical state data includes:

[0066] Step S201: Based on the analysis of the historical sensing data and the historical state data, determine the historical sensing parameters and historical load parameters at different time nodes within the same time period, and thereby construct several historical sensing parameter curves and historical load parameter curves. Denote the combination of the historical sensing parameter curves and historical load parameter curves within the same time period as a historical parameter curve group.

[0067] This step involves analyzing the historical sensing data and historical load data, aiming to identify the data at different time nodes within the same time period; through these data, historical sensing parameter curves and historical load parameter curves can be constructed, and these curves represent the operating state of the equipment within a specific time period.

[0068] Combine the historical sensing parameter curves and historical load parameter curves within the same time period together to form a historical parameter curve group, which can better understand the variation of different parameters over time.

[0069] Step S202: Perform grid-based dimensionality reduction on each historical sensing parameter curve and historical load parameter curve in the historical parameter curve group. The method of grid-based dimensionality reduction includes mapping the historical sensing parameter curve or historical load parameter curve onto a grid array template, and recording the reduced-dimension grids in the grid array that are mapped by the historical sensing parameter curve or historical load parameter curve, to obtain a grid array diagram, and denoting the combination of grid array diagrams corresponding to the historical parameter curve group as a grid array diagram group.

[0070] Grid-based dimensionality reduction is a data compression technique that converts high-dimensional data (such as historical parameter curves) into a lower-dimensional representation. In this step, the historical sensing parameter curves and historical load parameter curves are mapped onto a grid array template, and only the grids mapped by the curves are recorded, thereby obtaining a grid array diagram.

[0071] The grid array diagram group is composed of multiple grid array diagrams, and each grid array diagram corresponds to a historical parameter curve group. This can reduce the complexity of data processing while retaining key information.

[0072] Step S203: Calculate the average grid array diagram of the grid array diagrams of the same type within the grid array diagram group, and analyze the trend label of the corresponding grid array diagram in each grid array diagram group relative to the average grid array diagram. The method of judging the trend label includes analyzing the number of the first grids in the grid array diagram where the reduced-dimension grids mapped are higher than those in the average grid array diagram, and analyzing the number of the second grids in the grid array diagram where the reduced-dimension grids mapped are lower than those in the average grid array diagram, and determining the ratio of the number of grids of the first and second grids as the trend label.

[0073] In this step, calculate the average grid array diagram of the grid array diagrams of the same type within the grid array diagram group. This helps to identify typical operating patterns.

[0074] The trend label quantifies the trend of the grid array diagram relative to the average grid array diagram. By comparing the number of grids above the average value and the number of grids below the average value in the grid array diagram, a ratio can be obtained, and this ratio is used as the trend label, reflecting the relative change trend of the data.

[0075] Step S204: Set a number of grid quantity ratio intervals for the trend label of each type of grid array diagram, and denote the combination of the grid quantity ratio intervals corresponding to different types of grid array diagrams as a ratio interval classification group.

[0076] Set grid quantity ratio intervals for each type of grid array diagram, and combine these intervals to form a ratio interval classification group. Such classification helps to group grid array diagrams with similar trends together.

[0077] Step S205: Using the ratio interval classification group as the classification benchmark, analyze and classify different grid array diagram groups, conduct a complete comparison of the grid array diagram groups within each category set, and eliminate the duplicate grid array diagram groups.

[0078] Classify different grid array diagram groups based on the ratio interval classification group. After classification, conduct a complete comparison of the grid array diagram groups within each category and eliminate the duplicate grid array diagram groups.

[0079] Eliminating the duplicate grid array diagram groups is to remove redundant information, ensure that each grid array diagram group is unique, and thus improve the accuracy and efficiency of subsequent analysis.

[0080] Step S206: Recognize the historical parameter curve group corresponding to the grid array diagram group after elimination as the sub-historical state data that has passed the independence screening.

[0081] After eliminating the duplicate grid array diagram groups, the historical parameter curve group corresponding to the remaining grid array diagram groups is considered as the sub-historical state data that has passed the independence screening. These data can be used for subsequent simulation and prediction because they represent the unique characteristics of the device under different operating states.

[0082] In some embodiments disclosed in the present invention, the method for eliminating the duplicate grid array diagram groups includes:

[0083] Step S2051: Combine the grid array diagram groups within the same category set in pairs, analyze the similarity degree between the grid array diagram groups in each pair. If the similarity degree is greater than or equal to the preset value, randomly eliminate one of the grid array diagram groups.

[0084] In some embodiments disclosed in the present invention, the method for calculating the similarity degree between the grid array diagram groups includes:

[0085] Step S20511: Correspond one-to-one the grid arrays of the same type between the grid array diagram groups, analyze the reduced-dimensional grids between the corresponding grid arrays, determine the number of matching grids that match, and the number of non-matching grids that do not match.

[0086] Step S20512: Calculate the sum of the number of non-matching grids and the number of non-matching grids to obtain the total number of grids, and based on the ratio relationship between the number of non-matching grids and the total number of grids, as well as the continuity characteristics of the non-matching grids, determine the similarity degree between the grid array diagram groups.

[0087] Among them, the expression for calculating the similarity degree between the grid array diagram groups is:

[0088]

[0089] Among them, S is the similarity degree between the grid array diagram groups, and K i is the weight coefficient of the i-th type of grid array diagram between the grid array diagram groups, and S max is the preset maximum similarity degree, and x 1～i is the number of matching grids of the i-th type between the grid array diagram groups, and x 2～i is the number of non-matching grids of the i-th type between the grid array diagram groups, and β[i] is the continuity judgment function of the non-matching grids of the i-th type between the grid array diagram groups. Based on the continuous number of non-matching grids, a corresponding continuity correction and adjustment coefficient is generated, and n is the number of types of grid array diagrams between the grid array diagram groups.

[0090] Among them, the expression of the continuity judgment function is:

[0091]

[0092] Among them, L is the continuity conversion coefficient, α is the continuous number of non-matching grids, γ is the number of continuous sections of non-matching grids, and b is the continuity adjustment constant.

[0093] Step S300: Based on each sub-historical state data, configure state parameters for each virtual component structure in the virtual hydro-generator set structure to obtain the virtual dynamic performance of the virtual hydro-generator set structure.

[0094] In this step, the state parameters of each component in the virtual hydro-generator set are configured using the sub-historical state data; this means mapping the eigenvalue in the historical data to the corresponding virtual component, simulating the operating states of each component under different working conditions, so as to obtain the dynamic performance of the entire virtual hydro-generator set under different historical states.

[0095] Step S400: Obtain the real-time state data of the hydro-generator set (including real-time sensing data and real-time load data), and based on the real-time load data in the real-time state data, perform consistency screening on the historical state data to obtain the reference historical state data with the highest degree of consistency with the real-time state data, and use the virtual dynamic performance of the virtual hydro-generator set structure corresponding to the reference historical state data as the reference virtual dynamic performance.

[0096] Monitor the operating state of the hydro-generator set in real time to obtain the real-time state data including sensing data and load data. Then, compare the real-time load data with the historical load data, and screen out the historical state data that best matches the current real-time state. The dynamic performance of the virtual hydro-generator set corresponding to these data is used as the reference virtual dynamic performance.

[0097] In some embodiments disclosed by the present invention, the method for performing consistency screening on historical status data includes:

[0098] Step S401: Perform hourly analysis on the historical load data and real-time load data in the historical status data, and based on the analysis results, construct a real-time load curve and a historical load curve.

[0099] Step S402: Establish a time reference axis for the real-time load curve and the historical load curve, and set a number of time reference sections on the time reference axis. Align the real-time load curve, the historical load curve, and the time reference axis, and calculate the average value of the real-time load curve and the average value of the historical load curve mapped by each time reference section.

[0100] Step S403: Compare the average value of the real-time load curve and the average value of the historical load curve one by one. If the cumulative average value difference is greater than or equal to a preset value, replace the historical status data corresponding to the historical load curve until, under the scale of the time reference axis, the cumulative average value difference is still less than the preset value. Then, use the corresponding historical status data as the calculation object for the real-time load data, and calculate the degree of consistency.

[0101] In some embodiments disclosed by the present invention, the expression for calculating the degree of consistency is:

[0102]

[0103] where W is the degree of consistency, W max is the preset maximum degree of consistency, R is the difference degree conversion coefficient, Δh g is the cross-sectional area of the g-th intersection interval between the real-time load curve and the historical load curve, G is the total number of intersection intervals between the real-time load curve and the historical load curve, μ is the variance between the real-time load curve and the historical load curve, and c is the variance adjustment constant.

[0104] Step S500: Compare the real-time sensing data in the real-time status data with the reference virtual dynamic performance, and identify the virtual components with difference parameters greater than or equal to the preset value as abnormal states, and mark the virtual component structures in the virtual hydro-generator unit structure that are in abnormal states.

[0105] Compare the real-time sensing data with the reference virtual dynamic performance obtained in step S400. If there is a significant difference between the real-time state and the reference state of a certain virtual component, that is, the difference parameter is greater than or equal to the preset value, then this component is identified as an abnormal state. Mark these components in the virtual hydro-generator unit structure in an abnormal state for further analysis and processing.

[0106] Through the above steps, operators can promptly detect potential fault risks, take preventive measures, avoid equipment damage and unexpected shutdowns, thereby improving the safety and reliability of hydro-generator units.

[0107] In some embodiments disclosed by the present invention, there is also disclosed a fault warning system for hydro-generator units based on digital twins, including:

[0108] A first module, configured to obtain the structural information of the hydro-generator unit, and based on the structural information of the hydro-generator unit, construct a virtual component structure of the hydro-generator unit, combine the virtual component structures, and define the motion form of each virtual component structure and the motion range for the motion form, so as to obtain a virtual hydro-generator unit structure;

[0109] A second module, configured to obtain the historical state data of the hydro-generator unit, where the historical state data includes historical sensing data and historical load data, and perform independence screening on the data in the historical state data to obtain several sub-historical state data;

[0110] A third module, configured to configure state parameters for each virtual component structure in the virtual hydro-generator unit structure based on each sub-historical state data, so as to obtain the virtual dynamic performance of the virtual hydro-generator unit structure;

[0111] A fourth module, configured to obtain the real-time state data of the hydro-generator unit (including real-time sensing data and real-time load data), and based on the real-time load data in the real-time state data, perform conformity screening on the historical state data to obtain the reference historical state data with the highest degree of conformity with the real-time state data, and use the virtual dynamic performance of the virtual hydro-generator unit structure corresponding to the reference historical state data as the reference virtual dynamic performance;

[0112] A fifth module, configured to compare the real-time sensing data in the real-time state data with the reference virtual dynamic performance, and identify the virtual components with difference parameters greater than or equal to the preset value as abnormal states, and mark the virtual component structures in the virtual hydro-generator unit structure that are in abnormal states.

[0113] The present invention discloses a method and system for fault early warning of hydraulic generator sets based on digital twins, which relates to the technical field of fault analysis of hydraulic generator sets. Specifically, it discloses that state parameters are configured for each virtual component structure in the virtual structure of the hydraulic generator set, the historical state data is screened for consistency, and the reference historical state data with the highest degree of consistency with the real-time state data is obtained. The virtual dynamic performance of the virtual hydraulic generator set structure corresponding to the reference historical state data is used as the reference virtual dynamic performance. The real-time sensing data in the real-time state data is compared with the reference virtual dynamic performance, and based on the comparison result, the virtual component structures in the abnormal state on the virtual hydraulic generator set structure are marked. Through the above technical solutions, the present invention realizes the analysis of the real-time state data of the hydraulic generator set, and determines the components with abnormalities based on the analysis results. Compared with manual inspection, it not only has higher efficiency but also higher accuracy.

[0114] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered within the protection scope of the present invention.

Claims

1. A fault warning method for a hydraulic generator set based on digital twin, characterized in that, Including: Obtain the structural information of the water conservancy generating unit, and based on the structural information of the water conservancy generating unit, construct the virtual component structure of the water conservancy generating unit, combine the virtual component structures, and define the motion form of each virtual component structure and the motion range for the motion form to obtain the virtual water conservancy generating unit structure; Obtain the historical state data of the water conservancy generating unit, where the historical state data includes historical sensing data and historical load data, and perform independent screening on the data in the historical state data to obtain several sub-historical state data; Based on each sub-historical state data, configure the state parameters of each virtual component structure in the virtual water conservancy generating unit structure to obtain the virtual dynamic performance of the virtual water conservancy generating unit structure; Obtain the real-time state data of the water conservancy generating unit, and based on the real-time load data in the real-time state data, perform consistency screening on the historical state data to obtain the reference historical state data with the highest degree of consistency with the real-time state data, and use the virtual dynamic performance of the virtual water conservancy generating unit structure corresponding to the reference historical state data as the reference virtual dynamic performance; Compare the real-time sensing data in the real-time state data with the reference virtual dynamic performance, and identify the virtual components with the difference parameter greater than or equal to the preset value as abnormal states, and mark the virtual component structures in the virtual water conservancy generating unit structure that are in abnormal states.

2. The fault warning method for a hydraulic generator unit based on digital twin according to claim 1, wherein The method for constructing the virtual water conservancy generating unit structure includes: Construct a three-dimensional coordinate system, and configure different virtual component structures in the three-dimensional coordinate system for combination; Based on the motion range of the motion form of the virtual component structure, delimit the motion space in the three-dimensional coordinate system, and delimit the safe motion subspace and several levels of abnormal motion subspaces for the motion space. Based on the state data of the virtual component structure, drive the virtual component structure to move in the corresponding motion subspace.

3. The fault warning method for a hydraulic generator unit based on digital twin according to claim 1, wherein, The historical sensing data includes: turbine temperature data, generator temperature data, bearing temperature data, transformer temperature data, bearing vibration data, rotor vibration data, stator vibration data, turbine inlet and outlet pressure data, pipeline system pressure data, turbine flow data, axial displacement data of the rotor, voltage data, current data, lubricating system oil pressure data, lubricating system oil temperature data.

4. The fault warning method for a hydraulic generator set based on digital twin according to claim 1, wherein, The method for performing independent screening on the data in the historical state data includes: Based on the analysis of the historical sensing data and the historical state data, determine the historical sensing parameters and historical load parameters in the same time period and at different time nodes in the time period, and construct several historical sensing parameter curves and historical load parameter curves from this. Denote the combination of the historical sensing parameter curve and the historical load parameter curve in the same time period as the historical parameter curve group; Performing grid dimension reduction on each historical sensor parameter curve and historical load parameter curve in the historical parameter curve group, the grid dimension reduction method comprising mapping the historical sensor parameter curve or the historical load parameter curve into a grid array template, and recording the dimension reduction grids mapped by the historical sensor parameter curve or the historical load parameter curve in the grid array to obtain a grid array diagram, and recording the combination of the grid array diagrams corresponding to the historical parameter curve group as a grid array diagram group; Calculating the average square array graph of the square array graphs of the same type between the square array graph groups, and analyzing the trend labels of the corresponding square array graphs in each square array graph group relative to the average square array graph, the method for determining the trend label includes analyzing the number of first squares in which the dimensionality reduction squares mapped in the square array graph are higher than the dimensionality reduction squares mapped in the average square array graph, and analyzing the number of second squares in which the dimensionality reduction squares mapped in the square array graph are lower than the dimensionality reduction squares mapped in the average square array graph, and identifying the ratio of the first square number to the second square number as the trend label; A plurality of square quantity ratio intervals are set for the trend labels of each type of square array graph, and the combination of square quantity ratio intervals corresponding to different types of square array graphs is recorded as a ratio interval classification group; Using the ratio interval classification group as the classification basis, different grid array groups are analyzed and classified, and the grid array groups in each category set are completely compared, and repeated grid array groups are eliminated; The historical parameter curve group corresponding to the eliminated grid array graph group is identified as the sub-historical state data that has been screened for independence.

5. The fault warning method for a hydraulic generator unit based on digital twin according to claim 4, characterized in that, Methods for removing repeated grid array groups include: The square array graph groups belonging to the same category set are combined in pairs, and the similarity between the square array graph groups in each pair of combinations is analyzed. If the similarity is greater than or equal to a preset value, one of the square array graph groups is randomly eliminated.

6. The fault warning method for a hydraulic generator unit based on digital twin according to claim 5, characterized in that, Methods for calculating the similarity between grid array graph groups include: The same type of square array diagrams between the square array diagram groups are matched one by one, and the dimension reduction squares between the corresponding square array diagrams are analyzed to determine the number of matching squares and the number of non-matching squares; Calculate the number of non-matching squares and the sum of the number of non-matching squares to obtain the total number of squares, and determine the similarity between the square array graph groups based on the ratio of the number of non-matching squares to the total number of squares and the continuity characteristics of the non-matching squares; The expression for calculating the similarity between grid array graph groups is: Among them, S is the similarity degree between the square grid array diagram groups, and K i is the weight coefficient of the i-th type of square grid array diagram among the square grid array diagram groups, and S max is the preset maximum similarity degree, x 1～i is the number of matching squares of the i-th type among the square grid array diagram groups, and x 2～i is the number of non-matching squares of the i-th type among the square grid array diagram groups. β[i] is the continuity judgment function of the non-matching squares of the i-th type among the square grid array diagram groups. Based on the continuous number of non-matching squares, a corresponding continuity correction adjustment coefficient is generated. n is the number of types of square grid array diagrams among the square grid array diagram groups; Among them, the expression of the continuity judgment function is: Among them, L is the continuity conversion coefficient, α is the number of consecutive non-matching squares, γ is the number of consecutive sections of non-matching squares, and b is the continuity adjustment constant.

7. The fault warning method for a hydraulic generator set based on digital twin according to claim 1, wherein, Methods for matching historical status data include: Analyze the historical load data and real-time load data in the historical status data hour by hour, and construct the real-time load curve and the historical load curve based on the analysis results; A time reference axis is established for the real-time load curve and the historical load curve, and a number of time reference segments are set on the time reference axis, the real-time load curve, the historical load curve and the time reference axis are aligned, and the average value of the real-time load curve and the average value of the historical load curve mapped in each time reference segment are calculated; The average value of the real-time load curve is compared with the average value of the historical load curve one by one. If the cumulative difference in the average value is greater than or equal to the preset value, the historical status data corresponding to the historical load curve is replaced until the cumulative difference in the average value is still less than the preset value on the time reference axis scale. The corresponding historical status data is used as the calculation object for the real-time load data to calculate the degree of consistency.

8. The fault warning method for a hydraulic generator unit based on digital twin according to claim 7, wherein The expression for calculating the degree of fit is: Among them, W is the degree of fit, W max is the preset maximum matching degree, R is the difference degree conversion coefficient, Δh g is the intersection area of ​​the gth intersection interval between the real-time load curve and the historical load curve, G is the total number of intersection intervals between the real-time load curve and the historical load curve, μ is the variance between the real-time load curve and the historical load curve, and c is the variance adjustment constant.

9. A fault warning system for a hydraulic generator set based on digital twin, characterized in that, include: The first module is used to obtain the structural information of the hydropower generator set, and based on the structural information of the hydropower generator set, construct a virtual component structure of the hydropower generator set, combine the virtual component structures, and define the motion form of each virtual component structure and the motion range for the motion form to obtain the virtual hydropower generator set structure; The second module is used to obtain the historical status data of the hydropower generating set, the historical status data includes historical sensor data and historical load data, and perform independence screening on the data in the historical status data to obtain a plurality of sub-historical status data; The third module is used to configure the state parameters of each virtual component structure in the virtual hydropower generator set structure based on each sub-historical state data, so as to obtain the virtual dynamic performance of the virtual hydropower generator set structure; The fourth module is used to obtain the real-time status data of the hydropower generating set, and based on the real-time load data in the real-time status data, screen the historical status data for consistency, obtain the reference historical status data with the highest degree of consistency with the real-time status data, and use the virtual dynamic performance of the virtual hydropower generating set structure corresponding to the reference historical status data as the reference virtual dynamic performance; The fifth module is used to compare the real-time sensor data in the real-time status data with the reference virtual dynamic performance, and identify the virtual components whose difference parameters are greater than or equal to the preset values ​​as abnormal states, and mark the virtual component structures in abnormal states on the virtual hydropower generator set structure.

Images