A machine learning-based simulation and analysis system for power equipment

Abstract

This invention provides a machine learning-based power equipment simulation and analysis system, comprising a data acquisition module, a data processing module, a power equipment simulation module, and a fault analysis module. By integrating electro-thermal coupling modeling, orthogonal basis order reduction analysis, and modular ROM assembly, a rapid simulation model is constructed, significantly improving the modeling efficiency and response speed of power equipment under complex operating conditions. The simulation module introduces a GL-MLP-Trans model to achieve intelligent prediction and stability control of the simulation scheme. The fault analysis module employs a low-rank-Bi-LSTM model, combined with a spectral projection gain mechanism and a sub-modulus function approximation projection strategy, reducing model complexity while improving fault identification accuracy. This system is applicable to smart grids, power operation and maintenance, and industrial equipment condition monitoring scenarios, possessing high efficiency, robustness, and engineering adaptability.

Description

Technical Field

[0001] This invention relates to the field of power equipment simulation technology, and in particular to a power equipment simulation and analysis system based on machine learning. Background Technology

[0002] With the continuous expansion of power system scale and the increasing complexity of equipment operating environments, real-time perception of power equipment operating status and fault prediction have become crucial for ensuring the safe and reliable operation of the system. To achieve a comprehensive understanding and efficient management of power equipment operating behavior, power equipment simulation technology has gradually become an important tool in research and engineering practice. Currently, traditional power equipment simulation methods mainly rely on detailed finite element modeling and numerical solution processes, such as establishing electro-thermal coupling models to accurately simulate physical quantities like equipment temperature rise and current density. Although these methods have certain advantages in accuracy, their high computational complexity and long simulation time make them unsuitable for application scenarios with multiple operating conditions and strong real-time requirements in large-scale power systems. In recent years, artificial intelligence, especially machine learning technology... The rapid development of technology has provided a new path for the rapid simulation and intelligent fault analysis of power equipment. By introducing methods such as data-driven modeling, neural network structures, and reduced-order modeling, the operating status of power equipment can be efficiently predicted and evaluated while ensuring accuracy. However, existing machine learning-based simulation analysis methods still have many shortcomings: First, traditional machine learning methods have limited ability to integrate operating condition characteristics and equipment structural information, making it difficult to accurately model the dynamic behavior between complex components. Second, in the process of assembling reduced-order models, there is a lack of a unified optimization and control mechanism between simulation accuracy and model stability. Third, existing fault diagnosis models generally suffer from large parameter scales and significant redundancy, making it difficult to balance the generalization ability of modeling with the computational efficiency of operation. Summary of the Invention

[0003] This invention aims to overcome the shortcomings of existing power equipment simulation and analysis methods in terms of computational efficiency, modeling accuracy, and fault identification capabilities. It provides a machine learning-based power equipment simulation and analysis system that integrates electro-thermal coupled physical modeling, modular reduced-order modeling (ROM) technology, deep learning optimization mechanisms, and low-rank fault analysis methods to achieve rapid simulation and intelligent fault diagnosis of power equipment. Specifically, the power equipment simulation module proposes a rapid simulation method that integrates electro-thermal coupled modeling, orthogonal basis reduced-order analysis (POD), and modular ROM assembly. It also introduces a GL-MLP-Trans machine learning model to predict and optimize the simulation scheme. This model enhances feature representation capabilities through a multi-channel ghost node mechanism and combines feature spectrum direction-weighted lyap... The unov exponential diagnostic strategy achieves dynamic stability control, thereby improving the assembly accuracy and prediction robustness of ROM simulation. In fault analysis, this invention constructs a low-rank-Bi-LSTM neural network model. By introducing low-rank perturbations and spectral projection gain mechanisms into the LSTM network weights, it effectively achieves structural compression and performance improvement of the fault model. Specifically, a submodulus function approximation projection algorithm is used to filter key singular values, improving the fault identification accuracy and stability of the model in a finite parameter space. Overall, this invention organically combines physical modeling and machine learning to construct a power equipment simulation and analysis system that is fast-responding, structurally reconfigurable, and possesses adaptive diagnostic capabilities. It can be widely applied in scenarios such as smart grids, power operation and maintenance, and industrial equipment condition assessment.

[0004] This invention provides a machine learning-based power equipment simulation and analysis system, which includes a data acquisition module, a data processing module, a power equipment simulation module, and a fault analysis module.

[0005] The data acquisition module collects heterogeneous signal data in real time through electrical transformers, thermal sensors, vibration detectors, and intelligent switch quantity acquisition units deployed on the power equipment side. It then unifies and standardizes the heterogeneous signal data according to the IEC-61850 protocol, performs clock synchronization processing, and achieves millisecond-level time alignment. Simultaneously, it binds each data entry with a unique device ID, acquisition node code, and channel number, and encapsulates the processed data into a unified structure format to obtain a real-time data set. It also collects historical operating data and equipment model parameters, combining them with the real-time data set to construct power equipment operating data.

[0006] The data processing module filters, reduces noise, and selects features from the power equipment operation data to generate standardized power equipment data;

[0007] The power equipment simulation module uses a rapid power equipment simulation method to process standardized power equipment data and obtain a set of simulation results. The rapid power equipment simulation method is constructed by integrating electro-thermal coupling modeling, orthogonal basis order reduction (POD) analysis, componentized ROM model construction, and machine learning optimization mechanism.

[0008] The fault analysis module constructs a low-rank Bi-LSTM model, processes the simulation result set through the low-rank Bi-LSTM model, performs fault analysis, and generates fault analysis results.

[0009] Furthermore, the process of processing standardized power equipment data using a rapid power equipment simulation method to obtain a set of simulation results specifically includes the following steps:

[0010] Step S1: Extract the geometric structure, electrical boundaries and physical parameters of standardized power equipment data, establish an electro-thermal coupling simulation model, and perform multi-condition simulation by combining the Navier-Stokes equations and energy conservation equations to obtain the time-series temperature field and current density physical quantities, forming simulation snapshot data;

[0011] Step S2: Expand the simulation snapshot data according to the spatial dimension, construct the snapshot matrix, and perform singular value decomposition to extract dynamic features and construct a low-dimensional POD (orthogonal basis reduction decomposition) basis function space; based on the POD basis function space, combine the structural topology partitioning information in the standardized power equipment data to perform regional decomposition and functional module partitioning of the power equipment, generate a componentized ROM model library, and output ROM component meta-information and reduced-order description parameters from the componentized ROM model library;

[0012] Step S3: Set the target simulation condition parameters and introduce the GL-MLP-Trans model. The GL-MLP-Trans model optimizes the multi-task MLP-Transer model by introducing a multi-channel ghost node mechanism and a feature spectrum direction weighted-Lyapunov exponent diagnostic mechanism. The target simulation condition parameters, ROM component meta-information, and reduced-order description parameters are used to construct a condition feature vector set, which is then input into the GL-MLP-Trans model for training. This predicts the optimal ROM combination scheme and POD truncation dimension, generating an assembled ROM simulation model. The GL-MLP-Trans model includes a Transformer model and a multi-task MLP model.

[0013] Step S4: Combine the assembled ROM simulation model, perform reduced-order simulation, and use the Galerkin projection method to perform time-series evolution on the assembled ROM simulation model to quickly output dynamic response quantities, including temperature rise distribution, current density, and electrothermal coupling stress, to obtain the simulation response sequence; extract the peak temperature rise, maximum current density, voltage fluctuation amplitude, and energy consumption ratio of the simulation response sequence; structure these indicators according to the operating condition dimension and module region division to form a multi-dimensional response result set, and output the simulation result set.

[0014] Furthermore, the process of predicting the optimal ROM combination scheme and the POD truncation dimension specifically includes the following steps:

[0015] Step S31: Input the working condition feature vector set into the Transformer model for training. During the training process, a multi-channel ghost node mechanism is introduced to generate a joint feature vector of working condition and structure.

[0016] The multi-channel ghost node mechanism specifically includes: adding multi-channel ghost node vectors to the intermediate representation output by the Transformer model, performing dimensional expansion and dimensional perturbation, and using a regularization penalty term to control the strength of dimensional perturbation to prevent overfitting during training.

[0017] Step S32: Input the joint feature vector of working condition and structure into the multi-task MLP model for training, and output two branches. Branch 1 predicts whether each ROM component participates in the assembly, and branch 2 predicts the truncated dimension of the target POD. During the training process, a feature spectrum direction weighted-Lyapunov exponent diagnostic mechanism is introduced to dynamically evaluate the stability of the training state and adjust it in stages, so as to generate the optimal ROM combination scheme and POD truncated dimension.

[0018] Furthermore, the fault analysis module, in the process of generating fault analysis results, specifically includes the following steps:

[0019] Step E1: Establish a Bi-LSTM model, inject low-rank weight parameters into the weight matrix of the Bi-LSTM model, perform low-rank perturbation compensation on the main weights to achieve efficient parameter expression and gradient compression, construct a low-rank Bi-LSTM model, input the simulation result set into the low-rank Bi-LSTM model, capture the forward and backward evolution characteristics of the power equipment state, and obtain the initial low-rank weight matrix.

[0020] Step E2: Perform singular value decomposition on the initial low-rank weight matrix, extract the set of singular values, calculate the first and second gradients corresponding to the set of singular values, obtain the sensitivity and local curvature of each singular value to the performance of the low-rank-Bi-LSTM model, and form a singular value importance score.

[0021] Step E3: Based on the singular value importance score, define a set quadratic performance loss function, approximate the set quadratic performance loss function with a submodular function, and construct a spectral projection gain function. The spectral projection gain function has submodularity and monotonicity. Using the spectral projection gain function as the basis of the search space, introduce a global rank budget constraint, and use a greedy search algorithm to traverse the set of singular values ​​layer by layer, iteratively select the singular value component with the maximum marginal performance gain, and generate an optimized low-rank weight matrix.

[0022] Step E4: Re-inject the optimized low-rank weight matrix into the low-rank-Bi-LSTM model to complete the fault feature modeling and classification task and generate fault analysis results.

[0023] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0024] This invention constructs a rapid simulation method for power equipment by integrating electro-thermal coupling modeling, orthogonal basis order reduction analysis, and modular ROM assembly mechanism. This method enables rapid modeling and dynamic response prediction of power equipment states under complex operating environments, effectively improving the simulation efficiency and adaptability of the system under multiple operating conditions and loads. Compared to traditional methods relying on high-precision finite element simulation, this method significantly reduces computational complexity, enabling flexible assembly and rapid response of equipment-level models. In particular, the GL-MLP-Trans model introduced in this invention, through a multi-channel perturbation enhancement mechanism and a stability control strategy weighted by characteristic spectrum direction, enhances the robustness and generalization ability of the simulation model under structural changes and operating condition uncertainties, solving the problems of model rigidity and unstable errors in traditional simulation methods. This mechanism allows the simulation module to automatically adjust the model structure and order reduction parameters according to real-time operating conditions, achieving reconfigurable and rapidly deployable engineering simulation capabilities.

[0025] In terms of fault analysis, this invention proposes a Bi-LSTM fault diagnosis method based on low-rank perturbation and spectral projection gain mechanism optimization. By introducing singular value screening and structure compression mechanisms, the method effectively controls the model parameter scale and training resource consumption while ensuring modeling accuracy. This method can accurately capture the forward and backward evolution laws in the equipment's operating state, improving the system's sensitivity and identification ability to potential fault features. In particular, the spectral gain screening mechanism, combined with the submodulus function approximate projection algorithm, can dynamically select key feature paths under limited parameter budget, avoiding redundant modeling and overfitting problems, and enhancing the practicality and real-time response capability of the fault diagnosis model. Compared with traditional models based on rules or full features, this method improves the accuracy of judgment while significantly reducing computation and deployment overhead, meeting the requirements of lightweight and timeliness for edge computing and on-site diagnosis.

[0026] In summary, this invention constructs a power equipment simulation and analysis system that integrates rapid modeling, intelligent prediction, and efficient diagnosis. By synergistically integrating physical modeling and intelligent algorithms, it achieves rapid response, accurate evaluation, and stable operation support for power equipment under actual complex operating conditions. Compared with traditional technical means, this invention has achieved fundamental breakthroughs in many aspects, such as simulation efficiency, diagnostic accuracy, model flexibility, and resource consumption. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the modules of a machine learning-based power equipment simulation and analysis system proposed in this invention.

[0028] Figure 2 This is the convergence curve of the Lyapunov exponent with directional weighting of the characteristic spectrum proposed in Example 4.

[0029] Figure 2 In the diagram, the orange curve represents the estimated Lyapunov exponent, the blue dashed line represents the moving average, the light blue area represents the confidence interval, and the red dot and "peak" label indicate the location of the maximum Lyapunov value. The bottom line represents the training steps (0-200), and the left line represents the Lyapunov exponent value. Detailed Implementation

[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] Example 1, according to Figure 1 The present invention provides a power equipment simulation and analysis system based on machine learning, which includes a data acquisition module, a data processing module, a power equipment simulation module and a fault analysis module;

[0032] The data acquisition module collects heterogeneous signal data in real time through electrical transformers, thermal sensors, vibration detectors, and intelligent switch quantity acquisition units deployed on the power equipment side. It then unifies and standardizes the heterogeneous signal data according to the IEC-61850 protocol, performs clock synchronization processing, and achieves millisecond-level time alignment. Simultaneously, it binds each data entry with a unique device ID, acquisition node code, and channel number, and encapsulates the processed data into a unified structure format to obtain a real-time data set. It also collects historical operating data and equipment model parameters, combining them with the real-time data set to construct power equipment operating data.

[0033] The data processing module filters, reduces noise, and selects features from the power equipment operation data to generate standardized power equipment data;

[0034] The power equipment simulation module uses a rapid power equipment simulation method to process standardized power equipment data and obtain a set of simulation results. The rapid power equipment simulation method is constructed by integrating electro-thermal coupling modeling, orthogonal basis order reduction (POD) analysis, componentized ROM model construction, and machine learning optimization mechanism.

[0035] The fault analysis module constructs a low-rank Bi-LSTM model, processes the simulation result set through the low-rank Bi-LSTM model, performs fault analysis, and generates fault analysis results.

[0036] Example 2, based on Example 1, describes the process of processing standardized power equipment data using a rapid power equipment simulation method to obtain a set of simulation results. The specific steps include:

[0037] Step S1: Extract the geometric structure, electrical boundaries and physical parameters of standardized power equipment data, establish an electro-thermal coupling simulation model, and perform multi-condition simulation by combining the Navier-Stokes equations and energy conservation equations to obtain the time-series temperature field and current density physical quantities, forming simulation snapshot data;

[0038] Step S2: Expand the simulation snapshot data according to the spatial dimension, construct the snapshot matrix, and perform singular value decomposition to extract dynamic features and construct a low-dimensional POD (orthogonal basis reduction decomposition) basis function space; based on the POD basis function space, combine the structural topology partitioning information in the standardized power equipment data to perform regional decomposition and functional module partitioning of the power equipment, generate a componentized ROM model library, and output ROM component meta-information and reduced-order description parameters from the componentized ROM model library;

[0039] Step S3: Set the target simulation condition parameters and introduce the GL-MLP-Trans model. The GL-MLP-Trans model optimizes the multi-task MLP-Transer model by introducing a multi-channel ghost node mechanism and a feature spectrum direction weighted-Lyapunov exponent diagnostic mechanism. The target simulation condition parameters, ROM component meta-information, and reduced-order description parameters are used to construct a condition feature vector set, which is then input into the GL-MLP-Trans model for training. This predicts the optimal ROM combination scheme and POD truncation dimension, generating an assembled ROM simulation model. The GL-MLP-Trans model includes a Transformer model and a multi-task MLP model.

[0040] The target simulation parameters include: environmental parameters, electrical load parameters, equipment settings, and thermal fluid parameters.

[0041] Step S4: Combine the assembled ROM simulation model, perform reduced-order simulation, and use the Galerkin projection method to perform time-series evolution on the assembled ROM simulation model to quickly output dynamic response quantities, including temperature rise distribution, current density, and electrothermal coupling stress, to obtain the simulation response sequence; extract the peak temperature rise, maximum current density, voltage fluctuation amplitude, and energy consumption ratio of the simulation response sequence; structure these indicators according to the operating condition dimension and module region division to form a multi-dimensional response result set, and output the simulation result set.

[0042] Example 3, based on Example 1, describes the process of processing standardized power equipment data to obtain a set of simulation results, specifically including the following steps:

[0043] Step Q1: Extract the geometric structure, electrical boundaries and physical parameters of standardized power equipment data, establish an electro-thermal coupling simulation model, and perform multi-condition simulation by combining the Navier-Stokes equations and energy conservation equations to obtain the time-series temperature field and current density physical quantities, forming simulation snapshot data;

[0044] Step Q2: Expand the simulation snapshot data according to the spatial dimension, construct the snapshot matrix, and perform singular value decomposition to extract dynamic features and construct a low-dimensional POD (orthogonal basis reduction decomposition) basis function space; based on the POD basis function space, combine the structural topology partitioning information in the standardized power equipment data to perform regional decomposition and functional module partitioning of the power equipment, generate a componentized ROM model library, and output ROM component meta-information and reduced-order description parameters.

[0045] Step Q3: Introduce the Transformer model, collect the target simulation condition parameters, and construct the condition feature vector set by combining the target simulation condition parameters with the ROM component meta-information and reduced-order description parameters. Input the vector into the Transformer model for training, predict the optimal ROM combination scheme and POD truncation dimension, and generate the assembled ROM simulation model.

[0046] Step Q4: Combine the assembled ROM simulation model, perform reduced-order simulation, and use the Galerkin projection method to perform time-series evolution on the assembled ROM simulation model to quickly output dynamic response quantities, including temperature rise distribution, current density, and electrothermal coupling stress, to obtain the simulation response sequence; extract the peak temperature rise, maximum current density, voltage fluctuation amplitude, and energy consumption ratio of the simulation response sequence; structure these indicators according to the operating condition dimension and module region division to form a multi-dimensional response result set, and output the simulation result set.

[0047] Example 4, according to Figure 2This embodiment is based on Embodiment 2. In this embodiment, the process of predicting the optimal ROM combination scheme and the POD truncation dimension specifically includes the following steps:

[0048] Step S31: Input the working condition feature vector set into the Transformer model for training. During the training process, a multi-channel ghost node mechanism is introduced to accelerate convergence in the early stage of training and regulate convergence stability in the middle and late stages, generating a joint feature vector of working condition and structure.

[0049] The multi-channel ghost node mechanism specifically includes: adding multi-channel ghost node vectors to the intermediate representation output by the Transformer model to perform dimensional expansion and dimensional perturbation, and using a regularization penalty term to control the strength of dimensional perturbation to prevent overfitting during training; in the early stage of training, this mechanism guides the Transformer model to break through local saddle points, improves its ability to explore the input distribution, and accelerates the convergence process of the context representation; in the later stage of training, the intervention of ghost nodes is gradually weakened through a time decay strategy to achieve progressive convergence of the enhanced representation vector, thereby improving the robustness and generalization ability of the Transformer model in modeling the working condition feature vector set;

[0050] Multi-channel ghost node mechanism structure and modeling method:

[0051] (1) Multi-channel disturbance coordination:

[0052] ;

[0053] in, Indicates a time step. This represents the enhanced representation vector after adding ghost nodes. This represents the original intermediate representation vector. This represents a vector concatenation operation. Indicates the time decay factor. Indicates an index variable. This represents the total number of ghost node channels. This represents the dynamic aggregation weights calculated based on the context. Indicates the first Each ghost node vector is a learnable vector used to represent the perturbation dimension. It is initially composed of random noise or perturbation and gradually participates in the representation training.

[0054] (2) Regular expression penalty control:

[0055] ;

[0056] in, This represents the regularized loss for ghost nodes. Represents the regularization coefficient; Indicates the first ghost node vectors The square of the L2 norm;

[0057] The multi-channel ghost node mechanism enhances the diversity and structural sensitivity of perturbation representations; dynamic aggregation enables perturbation representations to have scene selectivity and semantic adaptability; time decay and regularization control work together to ensure that the model can quickly break through the saddle point in the early stage of training and gradually converge in the middle and late stages.

[0058] Step S32: Input the joint feature vector of working condition and structure into the multi-task MLP model for training. The model outputs two branches: branch 1 predicts whether each ROM component participates in assembly, and branch 2 predicts the truncated dimension of the target POD. During training, a feature spectrum direction-weighted Lyapunov exponent diagnostic mechanism is introduced to dynamically evaluate the stability of the training state and adjust it in stages, generating the optimal ROM combination scheme and POD truncated dimension. The formula used is as follows:

[0059] The formula for the characteristic spectrum orientation-weighted Lyapunov exponent is as follows:

[0060] ;

[0061] in, This represents the cumulative number of steps. Indicates the first The characteristic spectrum orientation-weighted Lyapunov exponent estimate of the step. This represents the index of the current feature value during training. This represents the total number of eigenvalues ​​in the Hessian matrix, i.e., the parameter dimension of the multi-task MLP model. Indicates the first Time of the first Weighting coefficients for each spectral direction; Indicates the first The first step of the Hessian matrix at time step One eigenvalue; Indicates the learning rate. Indicates the first Step training, along the first A logarithmic measure of convergence of perturbations in each characteristic direction;

[0062] In this embodiment, Figure 2 The image shows the convergence curve of the feature spectrum orientation-weighted Lyapunov exponent during training steps 0-200.

[0063] Figure 2 In the middle, the orange curve represents the estimated Lyapunov exponent, the blue dashed line represents the moving average, the light blue area represents the confidence interval, and the red dot and "peak" label represent the location of the maximum Lyapunov value.

[0064] The feature spectrum direction-weighted Lyapunov exponent diagnostic mechanism is based on the classic Lyapunov exponent estimation method, and introduces a weighted expression of the Hessian feature spectrum structure to enhance the ability to distinguish convergence states with different perturbation directions and the sensitivity of the direction.

[0065] Example 5, based on Example 4, describes the process by which the fault analysis module generates fault analysis results, specifically including the following steps:

[0066] Step E1: Establish a Bi-LSTM model, inject low-rank weight parameters into the weight matrix of the Bi-LSTM model, perform low-rank perturbation compensation on the main weights to achieve efficient parameter expression and gradient compression, construct a low-rank Bi-LSTM model, input the simulation result set into the low-rank Bi-LSTM model, capture the forward and backward evolution characteristics of the power equipment state, and obtain the initial low-rank weight matrix.

[0067] Step E2: Perform singular value decomposition on the initial low-rank weight matrix, extract the set of singular values, calculate the first and second gradients corresponding to the set of singular values, obtain the sensitivity and local curvature of each singular value to the performance of the low-rank-Bi-LSTM model, and form a singular value importance score.

[0068] Step E3: Based on the singular value importance score, define a set quadratic performance loss function, approximate the set quadratic performance loss function with a submodular function, and construct a spectral projection gain function. The spectral projection gain function has submodularity and monotonicity. Using the spectral projection gain function as the basis of the search space, introduce a global rank budget constraint, and use a greedy search algorithm to traverse the set of singular values ​​layer by layer, iteratively select the singular value component with the maximum marginal performance gain, and generate an optimized low-rank weight matrix.

[0069] Step E4: Re-inject the optimized low-rank weight matrix into the low-rank-Bi-LSTM model to complete the fault feature modeling and classification task and generate fault analysis results;

[0070] The submodular function approximation projection refers to extracting the local gain structure from the ensemble quadratic performance loss function using the first-order gradient term and second-order Hessian information, constructing a function form that satisfies the submodularity and monotonicity constraints, facilitating subset optimization using greedy search. The formula used is as follows:

[0071] Define a quadratic performance loss function for a set:

[0072] ;

[0073] in, Indicates the range of ranks The singular value vector selected in the middle, Denotes the quadratic performance loss function of the set. Represents the Hessian matrix. This indicates the transpose operation. Represents the gradient vector. Represents a second-order term. Represents a first-order term;

[0074] This function represents the overall performance loss, from the rank range. Within, by selecting a subset of singular values (Other singular values ​​are set to zero), construct the spectral projection gain function, and characterize the marginal contribution of different combinations of singular values ​​to the performance;

[0075] Spectral projection gain function formula:

[0076] ;

[0077] in, Represents the set of singular values Spectral projection gain function, Represents singular value vectors Selecting an index set The corresponding sub-vector, Describes the quadratic performance loss function of the set. Regarding singular value vectors The first-order gradient vector, express and The inner product, express transpose, Describes the quadratic performance loss function of the set. Regarding singularity subsets The local second-order near, Describes the quadratic performance loss function of the set. Regarding singular value vectors The second-order partial derivative matrix, i.e., the Hessian submatrix, This represents the linear loss term along the gradient descent direction. This represents a second-order approximation term.

[0078] Example 6, based on Example 4, in this example, the process of generating fault analysis results by the fault analysis module specifically includes the following steps: establishing a Bi-LSTM model, inputting the simulation result set into the Bi-LSTM model, and generating fault analysis results.

[0079] Example 7, this example is based on Example 6, in this example,

[0080] The power equipment simulation module introduces a rapid power equipment simulation method. This method is constructed by integrating electro-thermal coupling modeling, orthogonal basis order reduction (POD) analysis, componentized ROM model construction, and machine learning optimization mechanisms. The rapid simulation method processes standardized power equipment data to obtain a set of simulation results.

[0081] Table 1 obtained in this embodiment shows the optimal ROM combination scheme and POD truncation dimension:

[0082] Table 1

[0083] ;

[0084] Table 2 presents the simulation results:

[0085] Table 2

[0086] ;

[0087] The fault analysis module constructs a low-rank Bi-LSTM model, processes the simulation result set through the low-rank Bi-LSTM model, performs fault analysis, and generates fault analysis results.

[0088] Table 3 shows the fault analysis results:

[0089] Table 3

[0090] .

[0091] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A machine learning-based power equipment simulation and analysis system, comprising a data processing module, wherein the data processing module collects power equipment operation data, processes the power equipment operation data, and generates standardized power equipment data; characterized in that: The system also includes a power equipment simulation module and a fault analysis module; The power equipment simulation module uses a rapid power equipment simulation method to process standardized power equipment data and obtain a set of simulation results. The fault analysis module constructs a low-rank-Bi-LSTM model, processes the equipment simulation results using the low-rank-Bi-LSTM model to perform fault analysis, and generates fault analysis results. The process of obtaining a set of simulation results using a rapid simulation method for power equipment includes the following steps: Step S1: Process standardized power equipment data to generate simulation snapshot data; Step S2: Based on the simulation snapshot data, generate a componentized ROM model library. The componentized ROM model library outputs ROM component meta-information and reduced-order description parameters. Step S3: Set the target simulation condition parameters, introduce the GL-MLP-Trans model, construct the condition feature vector set by combining the target simulation condition parameters with the ROM component meta-information and reduced-order description parameters, input the condition feature vector set into the GL-MLP-Trans model for training, predict the optimal ROM combination scheme and POD truncation dimension, and generate the assembled ROM simulation model. Step S4: Combine the assembled ROM simulation model and output the simulation result set; The fault analysis module generates fault analysis results, and the process includes the following steps: Step E1: Build a Bi-LSTM model, inject low-rank weight parameters into the weight matrix of the Bi-LSTM model to construct a low-rank Bi-LSTM model, and input the simulation result set into the low-rank Bi-LSTM model to obtain the initial low-rank weight matrix. Step E2: Perform singular value decomposition on the initial low-rank weight matrix to obtain a set of singular values, forming a singular value importance score; Step E3: Based on the singular value importance score, define the set quadratic performance loss function, perform submodular function approximation projection on the set quadratic performance loss function, and construct the spectral projection gain function; using the spectral projection gain function as the basis of the search space, introduce the global rank budget constraint, and generate the optimized low-rank weight matrix by traversing the set of singular values ​​through a greedy search algorithm. Step E4: Re-inject the optimized low-rank weight matrix into the low-rank-Bi-LSTM model to generate fault analysis results.

2. The power equipment simulation and analysis system based on machine learning according to claim 1, characterized in that: The GL-MLP-Trans model includes the Transformer model and the multi-task MLP model.

3. The power equipment simulation and analysis system based on machine learning according to claim 2, characterized in that: The process of predicting the optimal ROM combination scheme and the POD truncation dimension specifically includes the following steps: Step S31: Input the working condition feature vector set into the Transformer model for training to generate a joint working condition-structure feature vector; Step S32: Input the joint feature vector of working condition and structure into the multi-task MLP model for training. During the training process, a feature spectrum direction weighted-Lyapunov exponent diagnostic mechanism is introduced to dynamically evaluate the stability of the training state and adjust it in stages, so as to generate the optimal ROM combination scheme and POD truncation dimension.

4. The power equipment simulation and analysis system based on machine learning according to claim 3, characterized in that: In step S31, a multi-channel ghost node mechanism is introduced during training. Multi-channel ghost node vectors are added to the intermediate representation output by the Transformer model to perform dimensional perturbation. A regularization penalty term is used to control the strength of the dimensional perturbation.

Images