Power equipment fault prediction method and system based on large model
By building a large model based on self-attention mechanism, the problem of inefficient calculation in power equipment failure prediction is solved, and efficient power equipment failure prediction is achieved.
Patent Information
- Application Number
- CN202510612996.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-13
- Publication Date
- 2025-08-26
AI Technical Summary
The existing power equipment fault prediction method based on deep learning models is complex when processing long-time series data, and the model structure is complex and the parameters are numerous, resulting in low computing efficiency.
A large model based on the self-attention mechanism is used to predict power equipment failures. By building a lightweight architecture and combining the Transformer model of the self-attention mechanism, it processes the operating status and environmental data of the power equipment to predict faults.
While maintaining feature capture accuracy, it greatly reduces the parameter scale and calculation complexity, improving the computing efficiency and accuracy of power equipment failure prediction.
Smart Images
Figure CN120541369A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power equipment fault prediction, and in particular to a large model-based power equipment fault prediction method and system. Background Art
[0002] The stable operation of power equipment is one of the core elements in ensuring the reliability and safety of power systems. Traditional power equipment fault prediction methods are mostly based on simple linear models or empirical formulas. However, with the expansion of power system scale and the increase in equipment operation complexity, traditional power equipment fault prediction methods have difficulty accurately characterizing the complex nonlinear characteristics and multi-factor coupling relationships during power equipment operation. Deep learning methods, such as RNN, LSTM, and GRU, have been gradually applied to power equipment fault prediction in recent years. They utilize their powerful nonlinear modeling capabilities and ability to learn complex data patterns to address the shortcomings of the above-mentioned traditional methods. However, the data generated during the operation of power equipment is mostly long-term series data. When processing this type of data, deep learning models have low computational efficiency due to the complex model structure, numerous parameters, and large amount of computation. Summary of the Invention
[0003] In response to the above-mentioned defects, the present invention proposes a method and system for predicting power equipment faults based on a large model, aiming to solve the problem of low computational efficiency in the existing power equipment fault prediction method based on a deep learning model, when the deep learning model processes the long time series data generated by the operation of the power equipment due to the complex model structure, numerous parameters and large amount of calculation.
[0004] To achieve this object, the present invention adopts the following technical solutions:
[0005] A method for predicting power equipment failure based on a large model includes the following steps:
[0006] Step S1: collecting operating status data and external environment data of the power equipment to form a sample data set, and setting a dynamic threshold value of the power equipment fault prediction value based on the sample data set;
[0007] Step S2: preprocessing the sample data set to obtain a preprocessed sample data set, and dividing the preprocessed sample data set into a training set and a test set;
[0008] Step S3: Build a large model based on the self-attention mechanism;
[0009] Step S4: Use the training set to train the large model based on the self-attention mechanism to obtain the trained large model;
[0010] Step S5: collecting real-time data of power equipment;
[0011] Step S6: input the real-time data of the power equipment into the trained large model for fault prediction, and output the power equipment fault prediction value;
[0012] Step S7: Determine whether the power equipment failure prediction value is greater than or equal to the dynamic threshold of the power equipment failure prediction value. If so, determine that the power equipment has a potential failure risk and generate an early warning signal; if not, determine that the power equipment does not have a potential failure risk.
[0013] Preferably, in step S1, a dynamic threshold value of the power equipment fault prediction value is set according to the sample data set, and the specific calculation formula is as follows:
[0014] T = μ ± λ·σ;
[0015] Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
[0016] Preferably, step S2 specifically includes the following sub-steps:
[0017] Step S21: Smoothing the data in the sample data set using a sliding average method to obtain smoothed data. The specific calculation formula is as follows:
[0018]
[0019] in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series;
[0020] Step S22: normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows:
[0021]
[0022] in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set;
[0023] Step S23: extract key features from the normalized data using a mean calculation method.
[0024] Preferably, in step S3, the mathematical expression of the large model based on the self-attention mechanism is as follows:
[0025]
[0026] Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; d k represents the dimension of the key; softmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
[0027] Preferably, the method further comprises the following steps:
[0028] Based on the power equipment fault prediction value and the test set, the mean square error between the power equipment fault prediction value and the actual value is calculated. The specific calculation formula is as follows:
[0029]
[0030] Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set;
[0031] Minimize the mean square error between the predicted and actual values of power equipment faults to optimize large models based on the self-attention mechanism.
[0032] Another aspect of the present application provides a large model-based power equipment fault prediction system, the system comprising:
[0033] The first acquisition module is used to collect operating status data and external environment data of the power equipment to form a sample data set;
[0034] A setting module, used to set a dynamic threshold value of a power equipment fault prediction value based on a sample data set;
[0035] A preprocessing module, used to preprocess the sample data set to obtain a preprocessed sample data set;
[0036] A data partitioning module is used to divide the preprocessed sample data set into a training set and a test set;
[0037] Building blocks for constructing large models based on self-attention mechanisms;
[0038] The model training module is used to train the large model based on the self-attention mechanism using the training set to obtain the trained large model;
[0039] The second acquisition module is used to collect real-time data of power equipment;
[0040] The fault prediction module is used to input the real-time data of the power equipment into the trained large model for fault prediction and output the power equipment fault prediction value;
[0041] A judgment module is used to judge whether the power equipment failure prediction value is greater than or equal to the dynamic threshold value of the power equipment failure prediction value. If so, it is determined that the power equipment has a potential failure risk and the generation module is executed; if not, it is determined that the power equipment does not have a potential failure risk;
[0042] The generation module is used to generate early warning signals.
[0043] Preferably, in the setting module, a dynamic threshold value of the power equipment fault prediction value is set according to the sample data set, and the specific calculation formula is as follows:
[0044] T = μ ± λ·σ;
[0045] Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
[0046] Preferably, the preprocessing module includes:
[0047] The smoothing submodule is used to smooth the data in the sample data set using the sliding average method to obtain the smoothed data. The specific calculation formula is as follows:
[0048]
[0049] in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series;
[0050] The normalization processing submodule is used to normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows:
[0051]
[0052] in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set;
[0053] The feature extraction submodule is used to extract key features from the normalized data using the mean calculation method.
[0054] Preferably, in the building module, the mathematical expression of the large model based on the self-attention mechanism is as follows:
[0055]
[0056] Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; d k represents the dimension of the key; softmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
[0057] Preferably, it also includes:
[0058] The calculation module is used to calculate the mean square error between the power equipment fault prediction value and the actual value based on the power equipment fault prediction value and the test set. The specific calculation formula is as follows:
[0059]
[0060] Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set;
[0061] The mean square error minimization module is used to minimize the mean square error between the predicted value and the actual value of the power equipment fault to optimize the large model based on the self-attention mechanism.
[0062] The technical solutions provided by the embodiments of the present application may have the following beneficial effects:
[0063] In this scheme, a large model based on the self-attention mechanism is constructed to predict whether there is a potential failure risk of power equipment. Compared with the existing power equipment fault prediction method based on deep learning models, the large model based on the self-attention mechanism in this scheme adopts a lightweight architecture design, which greatly reduces the parameter scale while maintaining feature capture accuracy, and significantly reduces the computational complexity, so that the model can still maintain high computational efficiency when processing long time series data. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 The present invention is a flowchart of the steps of a method for predicting power equipment failure based on a large model. DETAILED DESCRIPTION
[0065] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention.
[0066] A method for predicting power equipment failure based on a large model includes the following steps:
[0067] Step S1: collecting operating status data and external environment data of the power equipment to form a sample data set, and setting a dynamic threshold value of the power equipment fault prediction value based on the sample data set;
[0068] Step S2: preprocessing the sample data set to obtain a preprocessed sample data set, and dividing the preprocessed sample data set into a training set and a test set;
[0069] Step S3: Build a large model based on the self-attention mechanism;
[0070] Step S4: Use the training set to train the large model based on the self-attention mechanism to obtain the trained large model;
[0071] Step S5: collecting real-time data of power equipment;
[0072] Step S6: input the real-time data of the power equipment into the trained large model for fault prediction, and output the power equipment fault prediction value;
[0073] Step S7: Determine whether the power equipment failure prediction value is greater than or equal to the dynamic threshold of the power equipment failure prediction value. If so, determine that the power equipment has a potential failure risk and generate an early warning signal; if not, determine that the power equipment does not have a potential failure risk.
[0074] This solution is a power equipment fault prediction method based on a large model, such as Figure 1As shown, the first step is to collect the operating status data and external environment data of the power equipment to form a sample dataset. Based on the sample dataset, a dynamic threshold for the power equipment fault prediction value is set. In this embodiment, the operating status data of the power equipment includes current, voltage, and temperature, and the external environment data of the power equipment includes temperature, humidity, and wind pressure. Collecting the operating status data and external environment data of the power equipment provides a data foundation for subsequent large-scale model training. Setting a dynamic threshold for the power equipment fault prediction value facilitates subsequent power equipment fault diagnosis. The second step is to preprocess the sample dataset to obtain a preprocessed sample dataset, and then divide the preprocessed sample dataset into a training set and a test set. In this embodiment, preprocessing the sample dataset helps improve the data quality of the operating status data and external environment data of the power equipment. By dividing the preprocessed sample dataset into a training set and a test set, the division of the training set facilitates the subsequent training of the large-scale model based on the self-attention mechanism, and the division of the test set facilitates the subsequent verification of the large-scale model based on the self-attention mechanism. The third step is to construct a large model based on the self-attention mechanism. In this embodiment, the large model based on the self-attention mechanism is a Transformer model. This large model based on the self-attention mechanism can effectively capture long-term and short-term dependencies in power equipment operating data, achieving accurate modeling of complex nonlinear time series. By constructing a large model based on the self-attention mechanism, it is facilitated for subsequent use in power equipment fault prediction. The fourth step is to train the large model based on the self-attention mechanism using a training set to obtain a trained large model. In this embodiment, training the large model based on the self-attention mechanism helps improve the accuracy and stability of the large model in predicting power equipment faults. The fifth step is to collect real-time data from power equipment. In this embodiment, collecting real-time data from power equipment provides a data foundation for subsequent power equipment fault prediction. The sixth step is to input the real-time data from power equipment into the trained large model for fault prediction and output a power equipment fault prediction value. In this embodiment, by accurately inputting the real-time data from power equipment into the trained large model, the trained large model can effectively capture subtle changes and underlying patterns in the operating status of power equipment, achieving precise power equipment fault prediction. The seventh step is to determine whether the power equipment failure prediction value is greater than or equal to the dynamic threshold of the power equipment failure prediction value. If so, the power equipment is determined to have a potential failure risk and a warning signal is generated. If not, the power equipment is determined to have no potential failure risk. In this embodiment, by comparing the power equipment failure prediction value with its dynamic threshold in real time, it is possible to accurately determine whether the power equipment has a potential failure risk. This avoids the misjudgment or omission of risks caused by the traditional method due to the difficulty of fixed thresholds in adapting to changes in equipment operating status, greatly improving the accuracy of fault risk identification. When a potential failure risk of power equipment is detected, the generation of a warning signal helps prompt personnel to conduct inspections and repairs.
[0075] In this scheme, a large model based on the self-attention mechanism is constructed to predict whether there is a potential failure risk of power equipment. Compared with the existing power equipment fault prediction method based on deep learning models, the large model based on the self-attention mechanism in this scheme adopts a lightweight architecture design, which greatly reduces the parameter scale while maintaining feature capture accuracy, and significantly reduces the computational complexity, so that the model can still maintain high computational efficiency when processing long time series data.
[0076] Preferably, in step S1, a dynamic threshold value of the power equipment fault prediction value is set according to the sample data set, and the specific calculation formula is as follows:
[0077] T = μ ± λ·σ;
[0078] Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
[0079] In this embodiment, λ is set to 2. By adopting a dynamic threshold setting method based on the mean and standard deviation of a sample data set, the actual distribution characteristics of the power equipment fault prediction value can be more accurately reflected.
[0080] Preferably, step S2 specifically includes the following sub-steps:
[0081] Step S21: Smoothing the data in the sample data set using a sliding average method to obtain smoothed data. The specific calculation formula is as follows:
[0082]
[0083] in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series;
[0084] Step S22: normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows:
[0085]
[0086] in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set;
[0087] Step S23: extract key features from the normalized data using a mean calculation method.
[0088] In this embodiment, in step S21, the data in the sample data set is smoothed by using a sliding average method, which is beneficial to reducing the noise in the sample data set and highlighting the overall trend or periodicity of the data. In step S22, the smoothed data is normalized to unify the scale of the data and prevent numerical overflow or underflow caused by an excessively large data range. In step S23, the key features in the normalized data are extracted by using a mean calculation method, which is beneficial to reducing the computational complexity of subsequent large model training based on the self-attention mechanism.
[0089] Preferably, in step S3, the mathematical expression of the large model based on the self-attention mechanism is as follows:
[0090]
[0091] Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; d k represents the dimension of the key; sotmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
[0092] In this embodiment, by adopting a large model based on the self-attention mechanism, the long-term and short-term time series dependency characteristics contained in the power equipment operation data can be effectively captured, thereby achieving accurate mathematical modeling of complex nonlinear time series.
[0093] Preferably, the method further comprises the following steps:
[0094] According to the power equipment fault prediction value and the test set, the mean square error between the power equipment fault prediction value and the actual value is calculated. The specific calculation formula is as follows:
[0095]
[0096] Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set;
[0097] Minimize the mean square error between the predicted and actual values of power equipment faults to optimize large models based on the self-attention mechanism.
[0098] In this embodiment, by minimizing the mean square error between the predicted value and the actual value of the power equipment fault, it is helpful to reduce the risk of overfitting of the large model based on the self-attention mechanism, thereby improving the prediction accuracy and reliability of the large model based on the self-attention mechanism.
[0099] Another aspect of the present application provides a large model-based power equipment fault prediction system, the system comprising:
[0100] The first acquisition module is used to collect operating status data and external environment data of the power equipment to form a sample data set;
[0101] A setting module, used to set a dynamic threshold value of a power equipment fault prediction value based on a sample data set;
[0102] A preprocessing module, used to preprocess the sample data set to obtain a preprocessed sample data set;
[0103] A data partitioning module is used to divide the preprocessed sample data set into a training set and a test set;
[0104] Building blocks for constructing large models based on self-attention mechanisms;
[0105] The model training module is used to train the large model based on the self-attention mechanism using the training set to obtain the trained large model;
[0106] The second acquisition module is used to collect real-time data of power equipment;
[0107] The fault prediction module is used to input the real-time data of the power equipment into the trained large model for fault prediction and output the power equipment fault prediction value;
[0108] A judgment module is used to judge whether the power equipment failure prediction value is greater than or equal to the dynamic threshold value of the power equipment failure prediction value. If so, it is determined that the power equipment has a potential failure risk and the generation module is executed; if not, it is determined that the power equipment does not have a potential failure risk;
[0109] The generation module is used to generate early warning signals.
[0110] This solution proposes a large-scale model-based power equipment fault prediction system. This system uses a first acquisition module, a setting module, a preprocessing module, a data partitioning module, a construction module, a model training module, a second acquisition module, a fault prediction module, a judgment module, and a generation module to predict faulty power equipment. The large-scale model, based on the self-attention mechanism, employs a lightweight architecture, significantly reducing parameter size and computational complexity while maintaining feature capture accuracy. This allows the model to maintain high computational efficiency even when processing long-term data series.
[0111] Preferably, in the setting module, a dynamic threshold of the power equipment fault prediction value is set according to the sample data set, and the specific calculation formula is as follows:
[0112] T = μ ± λ·σ;
[0113] Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
[0114] In this embodiment, by adopting a dynamic threshold setting method based on the mean and standard deviation of a sample data set, the actual distribution characteristics of the power equipment fault prediction value can be more accurately reflected.
[0115] Preferably, the pre-processing module includes:
[0116] The smoothing submodule is used to smooth the data in the sample data set using the sliding average method to obtain the smoothed data. The specific calculation formula is as follows:
[0117]
[0118] in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series;
[0119] The normalization processing submodule is used to normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows:
[0120]
[0121] in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set;
[0122] The feature extraction submodule is used to extract key features from the normalized data using the mean calculation method.
[0123] In this embodiment, the provision of a smoothing submodule helps reduce noise in the sample dataset and highlight the overall trend or periodicity of the data. The provision of a normalization submodule helps unify the scale of the data and prevent numerical overflow or underflow caused by an overly large data range. The provision of a feature extraction submodule helps reduce the computational complexity of subsequent large model training based on the self-attention mechanism.
[0124] Preferably, in the building module, the mathematical expression of the large model based on the self-attention mechanism is as follows:
[0125]
[0126] Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; dk represents the dimension of the key; softmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
[0127] In this embodiment, by adopting a large model based on the self-attention mechanism, the long-term and short-term time series dependency characteristics contained in the power equipment operation data can be effectively captured, thereby achieving accurate mathematical modeling of complex nonlinear time series.
[0128] Preferably, it also includes:
[0129] The calculation module is used to calculate the mean square error between the power equipment fault prediction value and the actual value based on the power equipment fault prediction value and the test set. The specific calculation formula is as follows:
[0130]
[0131] Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set;
[0132] The mean square error minimization module is used to minimize the mean square error between the predicted value and the actual value of the power equipment fault to optimize the large model based on the self-attention mechanism.
[0133] In this embodiment, by setting up a calculation module and a mean square error minimization module, it is helpful to reduce the risk of overfitting of the large model based on the self-attention mechanism, thereby improving the prediction accuracy and reliability of the large model based on the self-attention mechanism.
[0134] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.
[0135] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.
Claims
1. A method for predicting power equipment failure based on a large model, characterized by: The following steps are involved: Step S1: collecting operating status data and external environment data of the power equipment to form a sample data set, and setting a dynamic threshold value of the power equipment fault prediction value based on the sample data set; Step S2: preprocessing the sample data set to obtain a preprocessed sample data set, and dividing the preprocessed sample data set into a training set and a test set; Step S3: Build a large model based on the self-attention mechanism; Step S4: Use the training set to train the large model based on the self-attention mechanism to obtain the trained large model; Step S5: collecting real-time data of power equipment; Step S6: input the real-time data of the power equipment into the trained large model for fault prediction, and output the power equipment fault prediction value; Step S7: Determine whether the power equipment failure prediction value is greater than or equal to the dynamic threshold of the power equipment failure prediction value. If so, determine that the power equipment has a potential failure risk and generate an early warning signal; if not, determine that the power equipment does not have a potential failure risk.
2. The method for predicting power equipment failure based on a large model according to claim 1, characterized in that: In step S1, a dynamic threshold value of the power equipment fault prediction value is set according to the sample data set. The specific calculation formula is as follows: T = μ ± λ·σ; Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
3. The method for predicting power equipment failure based on a large model according to claim 1, characterized in that: In step S2, the following sub-steps are specifically included: Step S21: Smoothing the data in the sample data set using a sliding average method to obtain smoothed data. The specific calculation formula is as follows: in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series; Step S22: normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows: in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; Step S23: extract key features from the normalized data using a mean calculation method.
4. The method for predicting power equipment failure based on a large model according to claim 1, characterized in that: In step S3, the mathematical expression of the large model based on the self-attention mechanism is as follows: Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; d k represents the dimension of the key; softmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
5. The method for predicting power equipment failure based on a large model according to claim 1, characterized in that: The following steps are also included: Based on the power equipment fault prediction value and the test set, the mean square error between the power equipment fault prediction value and the actual value is calculated. The specific calculation formula is as follows: Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set; Minimize the mean square error between the predicted and actual values of power equipment faults to optimize large models based on the self-attention mechanism.
6. A large-model-based power equipment fault prediction system, using the large-model-based power equipment fault prediction method according to any one of claims 1 to 5, characterized in that: The system comprises: The first acquisition module is used to collect operating status data and external environment data of the power equipment to form a sample data set; A setting module, used to set a dynamic threshold value of a power equipment fault prediction value based on a sample data set; A preprocessing module, used to preprocess the sample data set to obtain a preprocessed sample data set; A data partitioning module is used to divide the preprocessed sample data set into a training set and a test set; Building blocks for constructing large models based on self-attention mechanisms; The model training module is used to train the large model based on the self-attention mechanism using the training set to obtain the trained large model; The second acquisition module is used to collect real-time data of power equipment; The fault prediction module is used to input the real-time data of the power equipment into the trained large model for fault prediction and output the power equipment fault prediction value; A judgment module is used to judge whether the power equipment failure prediction value is greater than or equal to the dynamic threshold value of the power equipment failure prediction value. If so, it is determined that the power equipment has a potential failure risk and the generation module is executed; if not, it is determined that the power equipment does not have a potential failure risk; The generation module is used to generate early warning signals.
7. The large model-based power equipment fault prediction system according to claim 6, characterized in that: In the setting module, a dynamic threshold value of the power equipment fault prediction value is set according to the sample data set. The specific calculation formula is as follows: T = μ ± λ·σ; Where T represents the dynamic threshold of the power equipment fault prediction value; μ represents the average value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; and λ represents the threshold sensitivity adjustment factor.
8. The large model-based power equipment fault prediction system according to claim 6, characterized in that: The pre-processing module comprises: The smoothing submodule is used to smooth the data in the sample data set using the sliding average method to obtain the smoothed data. The specific calculation formula is as follows: in, represents the data after smoothing; x i Represents the i-th data in the sample data set; w represents the size of the sliding window; t represents the time step in the time series; The normalization processing submodule is used to normalize the smoothed data to obtain normalized data. The specific calculation formula is as follows: in, represents the normalized data; μ represents the mean value of the data in the sample data set; σ represents the standard deviation of the data in the sample data set; The feature extraction submodule is used to extract key features from the normalized data using the mean calculation method.
9. The large model-based power equipment fault prediction system according to claim 6, characterized in that: In the building block, the mathematical expression of the large model based on the self-attention mechanism is as follows: Where Q represents the query matrix; K represents the key matrix; V represents the value matrix; d k represents the dimension of the key; softmax(x) represents the normalization function; Attention(x) represents the function of the self-attention mechanism.
10. The large model-based power equipment fault prediction system according to claim 6, characterized in that: Also includes: The calculation module is used to calculate the mean square error between the power equipment fault prediction value and the actual value based on the power equipment fault prediction value and the test set. The specific calculation formula is as follows: Where MSE represents the mean square error between the predicted value and the actual value of the power equipment fault; i represents the actual value of the i-th power equipment fault; represents the predicted value of the i-th power equipment failure; N represents the total number of data in the test set; The mean square error minimization module is used to minimize the mean square error between the predicted value and the actual value of the power equipment fault to optimize the large model based on the self-attention mechanism.