Data-driven method for condition assessment and life prediction of power transformer in nuclear power plant
By constructing a multiphysics coupled simulation system and data mining methods, the problems of accuracy assessment and life prediction of power transformer condition in nuclear power plants were solved. Real-time assessment of equipment condition and fault prediction were achieved, maintenance costs were reduced, and stable operation and life extension research of equipment were supported.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIAN THERMAL POWER RES INST CO LTD
- Filing Date
- 2022-11-25
- Publication Date
- 2026-06-30
AI Technical Summary
Existing technologies make it difficult to accurately assess the condition of power transformers in nuclear power plants, resulting in high maintenance costs and difficulty in predicting their lifespan. Traditional testing methods may damage the equipment and are not suitable for on-site testing.
By constructing a multiphysics coupled simulation system, combining data mining and intelligent analysis, we collect actual and simulation data, calculate Mahalanobis distance and classify health levels, and use particle filtering and support vector machine methods for state assessment and lifetime prediction.
It enables accurate assessment of the condition and prediction of the remaining life of power transformers in nuclear power plants, improves the accuracy and real-time nature of the assessment, reduces maintenance costs, and supports stable operation and life extension research of equipment.
Smart Images
Figure CN115730452B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electrical engineering, specifically relating to a data-driven method for condition assessment and life prediction of nuclear power plant power transformers. Background Technology
[0002] Power transformers are one of the core and critical pieces of equipment in the electrical system of a nuclear power plant. They are complex in structure, expensive to manufacture, and replacement is extremely costly. Accurately assessing the operating condition of power transformers, detecting faults early, and optimizing operation and maintenance strategies can significantly reduce maintenance costs and minimize time and economic losses. Therefore, conducting condition assessment and life prediction research on nuclear power plant power transformers to ensure their safe and reliable operation throughout their lifespan is of great significance for maintaining the stability of nuclear power plant operations and the economic efficiency of power equipment investment.
[0003] Typically, the lifespan of nuclear power plant transformers is determined by the difficulty in replacing paper insulation systems. Currently, the condition assessment methods for nuclear power plant transformers are mainly divided into two categories: physicochemical parameter testing and electrical parameter testing. Physicochemical parameter testing methods involve detecting dissolved gases in the transformer oil, and periodically sampling and testing the insulation paper (board) inside the transformer, such as furfural testing and degree of polymerization testing. Electrical parameter testing methods obtain aging status results through partial discharge tests and dielectric response analysis. These testing methods have certain limitations in practical application; the testing process may cause irreversible damage to the test samples, making them unsuitable for on-site testing. Summary of the Invention
[0004] The purpose of this invention is to provide a data-driven method for condition assessment and life prediction of nuclear power plant power transformers. This method has the advantages of high accuracy and good real-time performance when assessing the condition of nuclear power plant power transformers, and can also predict the remaining life of the equipment.
[0005] This invention is achieved through the following technical solution:
[0006] A data-driven method for condition assessment and lifetime prediction of nuclear power plant transformers includes the following steps:
[0007] Step 1: Collect actual test, maintenance and operation data of power transformers in nuclear power plants, and combine them with multiphysics simulation data to construct a dataset for condition assessment of power transformers in nuclear power plants.
[0008] Step 2: Based on the large set of actual and simulation data obtained in Step 1, data mining methods are used to calculate the Mahalanobis distance between the aging data of the nuclear power plant's power transformers and the health benchmark model, and the equipment is divided into different health levels accordingly to achieve an accurate assessment of the equipment status.
[0009] Step 3: Through intelligent analysis of the current health status and operating data of the nuclear power plant's power transformers obtained in Step 2, the prediction of faults and remaining lifespan is achieved.
[0010] A further improvement of the present invention is that step 1 includes collecting various actual data during the testing, overhaul, and operation and maintenance of the power transformers of the nuclear power plant by conducting on-site surveys and data collection at the operating nuclear power plant.
[0011] A further improvement of this invention is that step 1 further includes constructing a multiphysics coupling simulation system for nuclear power plant power transformers, comprising the following steps:
[0012] Based on the specific task envelope, operating modes, and structural model, the governing equations, boundary conditions, initial conditions, and excitation sources for the electromagnetic field, temperature field, and stress-strain field within the solution area are set. Considering the dependence of all material parameters on the physical fields and the source effects between different physical fields, the boundary value problem of a single physical field is coupled with other physical fields to obtain a multi-physics boundary value problem that more closely approximates the actual physical process, i.e., a set of nonlinear coupled partial differential equations. Furthermore, by combining auxiliary differential equations, multi-layer nesting, and finite element techniques, the spatiotemporal multi-scale and parametric nonlinearity problems encountered in the multi-physics coupling solution process are overcome, and a multi-physics model of the nuclear power plant's power transformer ("electromagnetic-thermal-mechanical") is established. A direct-indirect coupling method is used to solve the multi-physics model, i.e., the strongly coupled part is solved by direct coupling, and the weakly coupled part is solved by indirect coupling. Based on the above multi-physics coupling simulation system of the nuclear power plant's power transformer, combined with the Model Caro numerical analysis method, the equipment state under different operating conditions and different aging stages are simulated, extracting physicochemical and electrical parameters related to health characteristics. Through extensive simulations, aging simulation data required for condition assessment is obtained.
[0013] A further improvement of the present invention is that, in step 2, the actual data is preprocessed to eliminate the influence of background noise and other superimposed signals from the operating conditions on the data to be processed.
[0014] A further improvement of the present invention is that, in step 2, the aging simulation data and the preprocessed actual data are decoupled, the Mahalanobis distance between the aging data and the health benchmark model is calculated, and the data are divided into different health levels through cluster analysis to obtain the equipment aging status assessment results.
[0015] A further improvement of the present invention is that, in step 2, the current operating status parameters of the nuclear power plant power transformer are input into the established multiphysics field coupling simulation system, and based on the simulation output results, the real-time assessment of the equipment insulation status can be achieved.
[0016] A further improvement of this invention lies in step 3, which involves intelligently analyzing the current health status and operating data of the nuclear power plant power transformer to predict its remaining lifespan and potential faults. Due to the influence of environmental and operational stress factors, the insulation degradation process of nuclear power plant power transformers exhibits nonlinearity and time-varying characteristics, resulting in complex and difficult-to-describe performance degradation patterns. Addressing the issues of data randomness and large cumulative errors during measurement, this invention uses multi-source heterogeneous data of various physicochemical and electrical characteristic parameters as a basis to intelligently analyze the insulation status of the nuclear power plant power transformer, thereby predicting its remaining lifespan and insulation faults.
[0017] A further improvement of this invention is that it employs particle filtering and support vector machine methods to perform intelligent analysis of the insulation status of power transformers in nuclear power plants.
[0018] The present invention has at least the following beneficial technical effects:
[0019] This invention focuses on nuclear power plant transformers and proposes a data-driven method for condition assessment and remaining life prediction. This research provides a new approach for equipment condition monitoring, aging mechanism analysis, life assessment, and fault prediction in nuclear power plant transformers, offering theoretical and technical support for stable transformer operation and serving aging analysis and life extension studies. Furthermore, this method has significant theoretical and engineering application value for the development and research of digital twins of nuclear power plant transformers, providing technical support for the digital transformation of nuclear power plants. This method has the following advantages:
[0020] First: Compared with the traditional single-physics simulation method, the multi-physics simulation system for nuclear power plant power transformers established by this invention has more accurate simulation results because it fully considers the coupling effect between different physical fields.
[0021] Second: Overcome the shortcomings of manual experience judgment in traditional equipment condition monitoring and operation and maintenance decision-making processes, and improve the accuracy and real-time performance of condition assessment;
[0022] Third, it can predict the remaining lifespan and failures of power transformers in nuclear power plants;
[0023] Fourth, the method proposed in this invention is universal and can be extended to the condition assessment and life prediction of other key electrical equipment in nuclear power plants. Attached Figure Description
[0024] Figure 1 This is a schematic diagram illustrating the implementation process of the method proposed in this invention.
[0025] Figure 2 Flowchart for Monte Carlo analysis.
[0026] Figure 3 System design diagram for condition assessment and life prediction of power transformers in nuclear power plants. Detailed Implementation
[0027] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0028] like Figure 1 As shown, the data-driven method for condition assessment and life prediction of nuclear power plant transformers provided by this invention includes the following steps:
[0029] Step 1: Collect actual test, maintenance and operation data of power transformers in nuclear power plants, and combine them with multiphysics simulation data to construct a dataset for condition assessment of power transformers in nuclear power plants.
[0030] Step 1-1: Collect various data on the testing, overhaul, and operation and maintenance of power transformers at nuclear power plants through on-site surveys and data collection at operating nuclear power plants.
[0031] Steps 1-2: Construct a multi-physics field coupled simulation system for nuclear power plant power transformers to obtain simulation data of nuclear power plant power transformers under different operating conditions.
[0032] First, based on the specific task envelope, operating modes, and structural model, the governing equations, boundary conditions, initial conditions, and excitation sources for the electromagnetic field, temperature field, stress-strain field, etc., within the solution region are reasonably set. Considering material parameters dependent on the field distribution and source effects between different physical fields, the boundary value problem of a single physical field is coupled with other physical fields to obtain a multi-physics boundary value problem that more closely approximates the actual physical process, i.e., a set of nonlinear coupled partial differential equations. Furthermore, by combining auxiliary differential equations, multi-level nesting, and finite element techniques, the spatiotemporal multi-scale and parametric nonlinearity problems commonly encountered in multi-physics coupling solutions are overcome, and a multi-physics model of the nuclear power plant's power transformer ("electromagnetic-thermal-mechanical") is established.
[0033] Secondly, addressing the issues of high computational cost of direct coupling and difficulty in convergence of indirect coupling in strongly coupled problems in traditional multiphysics coupling algorithms, this invention employs a direct-indirect coupling approach to solve the multiphysics model. Specifically, it uses direct coupling to solve the strongly coupled parts and indirect coupling to solve the weakly coupled parts. This maximizes the advantages of both coupling methods, ensuring both convergence and efficiency of the algorithm.
[0034] Finally, based on the aforementioned multiphysics coupling simulation system of nuclear power plant power transformers, and combined with the Monte Carlo analysis method, such as... Figure 2 As shown, a large number of simulation analyses with different values for multiple parameters are performed to obtain the actual measurable physical and chemical parameters and electrical parameters of the equipment under different insulation aging states. At the same time, aging simulation data of system state characteristic parameters under different operating conditions and different load types are obtained.
[0035] Step 2: Based on the large set of actual and simulation data obtained in Step 1, data mining methods are used to calculate the Mahalanobis distance between the aging data of the nuclear power plant's power transformers and the health benchmark model, and the equipment is divided into different health levels to achieve an accurate assessment of the equipment status.
[0036] To address issues such as noise interference or superimposed signals from other operating conditions in online monitoring data and historical test and detection data, preprocessing of signals obtained during testing, maintenance, and operation is necessary. Using normal operation signals of key characteristic parameters under different operating conditions as training data, an intrinsic regression model of the normal signal is established to obtain relevant parameters and train a neural network model. Next, the residuals of the original intrinsic regression model and the residuals after neural network processing are extracted from the signal to be processed. The difference between the two residuals is used to characterize the operating status of the nuclear power plant's power transformer, thereby eliminating the influence of background noise and superimposed signals from other operating conditions on the signal to be processed.
[0037] The input and output formats are set, and data mining is performed on the simulation data and preprocessed monitoring data to achieve a state assessment of the power transformers in nuclear power plants. First, the original aging simulation data obtained from the Monte Carlo simulation analysis of the nuclear power plant power transformers is decoupled from the preprocessed actual test, maintenance, and operation data. After decoupling, the Mahalanobis distance between the decoupled aging data and the health baseline model is calculated, and cluster analysis is used to classify them into different health levels. Finally, a neural network model is trained using the aging dataset to obtain the final result of the system aging state assessment.
[0038] In addition, the current operating status parameters of the nuclear power plant's power transformer can be input into the multiphysics simulation system established by this invention. Based on the simulation output results, real-time evaluation of the equipment's operating status can be achieved.
[0039] Step 3: Through intelligent analysis of the current health status and operating data of the nuclear power plant's power transformers obtained in Step 2, the prediction of faults and remaining lifespan is achieved.
[0040] To address the issues of noise interference and large cumulative errors, this invention introduces spherical volume particle filtering to solve the state-space model and adaptively updates the model parameters, thereby tracking the performance degradation trend and establishing a particle filtering-based method for remaining lifetime and fault prediction.
[0041] To address the issues of limited effective data and small sample sizes that may exist in actual prediction processes, this invention employs the support vector machine method, which has good generalization capabilities, to construct the optimal decision function in space. By mapping the input sample vector factors to a high-dimensional feature space through a pre-selected nonlinear mapping, the nonlinear mapping relationship between the dependent and independent variables is sought in this high-dimensional space, ultimately achieving the prediction of remaining lifetime and insulation faults.
[0042] Step 4, System design for condition assessment and life prediction of nuclear power plant power transformers, such as... Figure 3 As shown, the system design includes a data module, a model module, and a software application module. The data module establishes a database for the equipment, containing various test, maintenance, and operation data, as well as simulation data obtained from the model module. The model module establishes a multi-physics coupled simulation system for the nuclear power plant's power transformers. This allows for simulation analysis of the transformer's operating status under different loads and operating conditions, and also enables real-time simulation of the transformer's operating status by inputting the equipment's current operating data. The software application module uses input algorithm software programs to perform status assessment, fault diagnosis, life prediction, and maintenance decisions for the nuclear power plant's power transformers. It also issues operation commands to the data and model modules based on the needs of assessment, prediction, and decision-making.
[0043] In summary, this invention proposes a data-driven method for the condition assessment and lifespan prediction of nuclear power plant power transformers. First, a multi-physics coupled simulation system for nuclear power plant power transformers is constructed. Combined with Monte Carlo analysis, the physicochemical and electrical characteristic parameters of the transformers under different operating conditions and aging stages are obtained. Additionally, testing, inspection, and operation and maintenance data of the nuclear power plant power transformers are collected. Next, data mining methods are used to classify the equipment into different health levels, enabling condition and aging assessments. Furthermore, through intelligent analysis of health status and operational data, the remaining lifespan of the nuclear power plant power transformers is predicted, and optimized operation and maintenance measures are proposed. This method effectively overcomes the shortcomings of traditional equipment condition monitoring and operation and maintenance decision-making based on human experience, improving the accuracy and real-time performance of equipment condition assessment and enabling the prediction of remaining lifespan and faults. This method provides theoretical and technical support for the stable operation of nuclear power plant power transformers and serves research on aging analysis and lifespan extension of nuclear power plant power transformers. Simultaneously, this method has significant theoretical and engineering application value for the development and research of digital twins of nuclear power plant power transformers, effectively promoting the digital transformation of nuclear power plants.
[0044] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.
Claims
1. A data-driven method for condition assessment and life prediction of nuclear power plant power transformers, characterized in that, Includes the following steps: Step 1 involves collecting actual test, maintenance, and operation data of nuclear power plant power transformers, and combining this with multiphysics simulation data to construct a dataset for condition assessment of nuclear power plant power transformers. This also includes building a multiphysics coupled simulation system for nuclear power plant power transformers, including: setting the governing equations, boundary conditions, initial conditions, and excitation sources for the electromagnetic field, temperature field, and stress-strain field within the solution area based on the specific task envelope, operating modes, and structural model; considering the dependence of all material parameters on the physical fields and the source effects between different physical fields; and coupling the boundary value problem of a single physical field with other physical fields to obtain a multiphysics boundary value problem that more closely approximates the actual physical process, i.e., a set of nonlinear coupled partial differential equations; further... This paper proposes a multi-physics model of nuclear power plant power transformers, combining auxiliary differential equations, multi-layer nesting, and finite element techniques to overcome the spatiotemporal multi-scale and parametric nonlinearity problems encountered in the multi-physics coupling solution process. A direct-indirect coupling approach is used to solve the multi-physics model, with direct coupling solving for strongly coupled parts and indirect coupling solving for weakly coupled parts. Based on this multi-physics coupling simulation system, and combined with Monte Carlo numerical analysis, the paper simulates the equipment status under different operating conditions and different aging stages, extracting physicochemical and electrical parameters related to health characteristics. Through extensive simulations, aging simulation data required for condition assessment is obtained. Step 2: Based on the large set of actual and simulation data obtained in Step 1, the aging simulation data and the preprocessed actual data are decoupled. Data mining methods are used to calculate the Mahalanobis distance between the aging data of the nuclear power plant power transformer and the health benchmark model. Cluster analysis is then used to divide the data into different health levels to obtain the equipment aging status assessment results. Step 3: Through intelligent analysis of the current health status and operating data of the nuclear power plant power transformer obtained in Step 2, the prediction of faults and remaining lifespan is achieved. Specifically, through intelligent analysis of the current health status and operating data of the nuclear power plant power transformer, the prediction of remaining lifespan and faults is achieved. Due to the influence of environmental stress and working stress factors, the insulation degradation process of the nuclear power plant power transformer is nonlinear and time-varying, and the performance degradation law is relatively complex and difficult to describe accurately. In view of the problems of data randomness and large cumulative error in the measurement process, based on multi-source heterogeneous data of various physicochemical and electrical characteristic parameters, the insulation status of the nuclear power plant power transformer is intelligently analyzed to predict the remaining lifespan and insulation faults of the nuclear power plant power transformer. The particle filtering method and the support vector machine method are used to perform intelligent analysis on the insulation status of power transformers in nuclear power plants.
2. The data-driven method for condition assessment and life prediction of nuclear power plant power transformers according to claim 1, characterized in that, Step 1 includes collecting various actual data during the testing, overhaul, and operation and maintenance of nuclear power plant power transformers by conducting on-site surveys and data collection at operating nuclear power plants.
3. The data-driven method for condition assessment and life prediction of nuclear power plant transformers according to claim 1, characterized in that, In step 2, the actual data is preprocessed to eliminate the influence of background noise and other superimposed signals from the operating conditions on the data to be processed.
4. The data-driven method for condition assessment and life prediction of nuclear power plant power transformers according to claim 1, characterized in that, In step 2, the current operating status parameters of the nuclear power plant's power transformer are input into the established multiphysics coupling simulation system. Based on the simulation output results, the real-time assessment of the equipment's insulation status can be achieved.