Power generation system fault auxiliary decision generation method and system based on large model

Abstract

The invention relates to the technical field of power systems, in particular to a power generation system fault auxiliary decision generation method and system based on a large model, and the method and system sequentially obtain a current detection information set of a power generation system, and pre-process the current detection information set to generate a time sequence data set; generating first prediction information and a residual sequence based on a preset ARIMA model according to the time sequence data set; inputting the residual sequence into a preset LSTM network to generate second prediction information; generating initial fault information according to the combination of the first prediction information and the second prediction information; inputting the preliminary fault detection information into a pre-trained Transform model for deep feature extraction and context association analysis, generating auxiliary decision information, and sending the auxiliary decision information to a manager; therefore, the accuracy of fault detection of the power generation system is improved, the intelligent level of auxiliary decision making is enhanced, and the operation stability and management efficiency of the system are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and particularly to a method and system for generating fault - assisted decision - making for a power generation system based on a large model. Background Art

[0002] With the continuous expansion of the scale of photovoltaic power generation systems and the increasing complexity of the operating environment, the stability and reliability of photovoltaic systems are particularly important. During the actual operation of a photovoltaic power generation system, the timely detection and accurate location of faults are crucial for ensuring the continuity of power supply and the safety of the system. However, the existing fault detection and decision - support solutions have the following problems in photovoltaic power generation systems.

[0003] First, the existing technologies are insufficient in data processing and feature extraction. Traditional photovoltaic fault detection methods often rely on simple statistical analysis or rule - based discrimination, and it is difficult to effectively process the large amount of real - time monitoring data from photovoltaic systems. These methods have limited capabilities in extracting complex time - series features and capturing potential patterns in the data, resulting in low accuracy and reliability of fault prediction.

[0004] Second, the existing decision - support systems are imperfect. Even if the existing methods can detect faults, when the existing auxiliary decision - making systems generate maintenance strategies and handling suggestions, they often lack the ability of in - depth feature extraction and context - related analysis. This leads to the generated decision information lacking pertinence and effectiveness, being unable to provide reliable decision - making support for managers, and thus affecting the efficiency and effect of fault handling.

[0005] Therefore, it is necessary to improve the existing technologies to solve the above problems. Summary of the Invention

[0006] Based on this, in view of the above - mentioned technical problems, it is necessary to provide a method and system for generating fault - assisted decision - making for a power generation system based on a large model to solve the technical problems of limited data processing and feature extraction capabilities and imperfect decision - support systems existing in the prior art.

[0007] The present invention provides a method for generating fault - assisted decision - making for a power generation system based on a large model, including:

[0008] Obtaining the current detection information set of the power generation system, and pre - processing the current detection information set to generate a time - series data set;

[0009] Generating first prediction information and a residual sequence based on the time - series data set according to a preset ARIMA model;

[0010] Inputting the residual sequence into a preset LSTM network to generate second prediction information;

[0011] Generate preliminary fault information by combining the first prediction information and the second prediction information;

[0012] Input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

[0013] As a preferred solution, generate the first prediction information and the residual sequence based on the time series data set based on a preset ARIMA model, including:

[0014] Select target parameters according to a preset historical fault database and expert experience database;

[0015] Traverse the time series data set to obtain the operating parameter sequences related to the target parameters, and generate the target operating parameter sequence according to each relevant operating parameter sequence;

[0016] Align and combine each target operating parameter sequence according to the time stamp to generate a target sequence, and input the target sequence into a preset ARIMA model to generate the first prediction information;

[0017] Calculate the prediction error according to the target sequence and the first prediction information, and generate a residual sequence.

[0018] As a preferred solution, input the residual sequence into a preset LSTM network to generate the second prediction information, including:

[0019] Extract time-dependent features from the residual sequence through a sliding window algorithm, and generate an input feature sequence for the residual sequence based on a set time step;

[0020] Extract statistical features, periodic features, and frequency domain features from the input feature sequence, and fuse the time-dependent features, the statistical features, the periodic features, and the frequency domain features to generate an enhanced feature sequence;

[0021] Input the enhanced feature sequence into a preset LSTM network for deep modeling to generate the first non-linear prediction information;

[0022] Perform short-term and medium-term non-linear prediction and trend analysis on the first non-linear prediction information based on a multi-step prediction algorithm to generate the second non-linear prediction information;

[0023] Filter and correct the second non-linear prediction information based on the expert experience database, and generate the second prediction information.

[0024] As a preferred solution, generate preliminary fault information by combining the first prediction information and the second prediction information, including:

[0025] Align the timestamps of the first prediction information and the second prediction information to generate an aligned prediction information set;

[0026] Based on mutual information measurement, perform differential evaluation and analysis on the first prediction information and the second prediction information within the same time interval, and generate a differential analysis index;

[0027] Generate a weight parameter according to the differential analysis index, and generate comprehensive prediction information according to the first prediction information, the second prediction information and the weight parameter;

[0028] Based on a pattern matching algorithm, generate fault matching information according to the comprehensive prediction information, the historical fault database and the expert experience database, wherein the fault matching information includes a historical fault identification flag and an expert rule flag;

[0029] Generate preliminary fault information based on the fault matching information using the Naive Bayes algorithm, wherein the preliminary fault information includes a historical fault matching flag, a fault type, a fault probability and a fault severity.

[0030] As a preferred solution, input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel, including:

[0031] Retrieve the corresponding fault case text information from the historical fault database according to the historical fault matching flag, query the corresponding processing suggestion text from the expert experience database according to the expert rule flag, and perform vectorization representation on the fault case text information and the processing suggestion text based on a text encoding algorithm to generate context embedding data;

[0032] Perform feature splicing and weighted integration according to the preliminary fault detection information and the context embedding data to generate a fused feature representation;

[0033] Input the fused feature representation into a pre-trained Transformer model, perform deep feature extraction and context correlation analysis on the fused feature through a multi-head attention mechanism, and generate a deep correlation feature;

[0034] Generate auxiliary decision-making candidate information according to the deep correlation feature and the expert experience database;

[0035] Perform result formatting and optimal screening on the auxiliary decision-making candidate information to generate auxiliary decision-making information, wherein the auxiliary decision-making information includes an executable maintenance strategy, a maintenance timing suggestion and a resource allocation plan;

[0036] Send the auxiliary decision-making information to the management personnel and obtain feedback information, and store the feedback information in the historical fault database and the expert experience database.

[0037] As a preferred solution, the ARIMA model is established based on the following steps:

[0038] Obtain the conventional detection information set of the power generation system, generate a conventional time series data set according to the conventional detection information set, and perform multi-granularity partitioning on the conventional time series data set to generate a modeling sequence and a verification sequence;

[0039] Decompose the modeling sequence through variational mode decomposition to generate multiple intrinsic mode components, establish corresponding initial ARIMA sub-models according to each intrinsic mode component, and perform iterative search and update on the parameters of the initial ARIMA sub-models through the Bayesian optimization algorithm to generate ARIMA sub-models;

[0040] Based on the ARIMA sub-model, use the fixed-length sliding window algorithm to perform rolling prediction on the verification sequence to generate a first prediction result;

[0041] Compare the first prediction result with the true value of the verification sequence to generate a residual sequence through residual refinement, and optimize the ARIMA sub-model according to the residual sequence to generate an optimized ARIMA sub-model;

[0042] Quantify the performance of the optimized ARIMA sub-model according to the residual sequence and generate an error value, and the error value is obtained based on the following formula:

[0043]

[0044] where E is the error value, N is the number of time steps of the verification sequence, t is the ordinal number of the time step, c t is the weight factor corresponding to the time step t, y t is the true value of the verification sequence at the time step t, y′ t is the model prediction value of the verification sequence at the time step t, and ε is a preset positive constant;

[0045] Generate sub-model weight parameters according to the error value, the historical fault database and the expert experience database, and construct the ARIMA model according to the ARIMA sub-model and the sub-model weight parameters.

[0046] As a preferred solution, obtain the current detection information set of the power generation system, and preprocess the current detection information set to generate a time series data set, including:

[0047] Obtain the current detection information set of the power generation system, and perform data fusion preprocessing on the current detection information set to generate a fusion information set, where the detection information set includes environmental detection information, system operation information, and status detection information;

[0048] Adopt the locally weighted regression interpolation algorithm for the fusion information set to adaptively fill in the missing values and generate a completed information set, and perform fitting correction on the completed information set through the random forest algorithm to generate a corrected information set;

[0049] Based on the z-score method, perform anomaly screening on the corrected information set to generate an outlier set, perform anomaly screening on the outlier set through the isolation forest algorithm to generate a confirmed outlier set, and generate a screening information set based on the locally weighted regression interpolation algorithm according to the completed detection information set and the confirmed outlier set;

[0050] Obtain the time series data of each detection feature in the screening information set, calculate the median and median absolute deviation value of each feature, and perform normalization processing on the screening information set according to the absolute deviation value to generate a normalized information set;

[0051] Based on the z-score method, standardize the normalized information set to generate a standard information set, and perform feature engineering processing on the standard information set to generate a feature extraction information set, where the feature extraction information set includes statistical features, periodic features, and frequency domain features.

[0052] The present invention also provides a power generation system fault auxiliary decision-making generation system based on a large model, including:

[0053] A time series data generation module, configured to obtain the current detection information set of the power generation system and perform preprocessing on the current detection information set to generate a time series data set;

[0054] A first prediction information generation module, configured to generate first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set;

[0055] A second prediction information generation module, configured to input the residual sequence into a preset LSTM network to generate second prediction information;

[0056] A preliminary fault information generation module, configured to generate preliminary fault information by combining the first prediction information and the second prediction information;

[0057] An auxiliary decision-making information generation module, configured to input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

[0058] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for generating an auxiliary decision on power generation system faults based on a large model are implemented.

[0059] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating an auxiliary decision on power generation system faults based on a large model are implemented.

[0060] The technical effects achieved by the present invention are as follows:

[0061] The above-mentioned method and system for generating an auxiliary decision on power generation system faults based on a large model sequentially obtain the current detection information set of the power generation system, preprocess the current detection information set to generate a time series data set; generate first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set; input the residual sequence into the preset LSTM network to generate second prediction information; generate preliminary fault information by combining the first prediction information and the second prediction information; input the preliminary fault detection information into the pre-trained Transformer model for in-depth feature extraction and context correlation analysis and generate auxiliary decision information, and send the auxiliary decision information to the management personnel; thereby achieving the improvement of the accuracy of power generation system fault detection, enhancing the intelligent level of auxiliary decision-making, significantly improving the operation stability and management efficiency of the system, and solving the key problems such as low prediction accuracy and limited decision support existing in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic flow chart of the method for generating an auxiliary decision on power generation system faults based on a large model in an embodiment;

[0063] Figure 2 It is a structural block diagram of the system for generating an auxiliary decision on power generation system faults based on a large model in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] In one embodiment, a terminal is provided. The terminal acquires the current detection information set of the power generation system, preprocesses the current detection information set to generate a time series data set; generates first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set; inputs the residual sequence into a preset LSTM network to generate second prediction information; generates preliminary fault information by combining the first prediction information and the second prediction information; inputs the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis to generate auxiliary decision-making information, and sends the auxiliary decision-making information to the management personnel.

[0066] The terminal may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices.

[0067] In this embodiment, as Figure 1 shown, a method for generating auxiliary decision-making for power generation system faults based on a large model is provided, which is particularly applicable to photovoltaic power generation systems. The method includes:

[0068] Step S100: Acquire the current detection information set of the power generation system, and preprocess the current detection information set to generate a time series data set.

[0069] In this step, the photovoltaic power generation system monitors devices such as photovoltaic modules, inverters, transformers, and junction boxes in real time, collects operating parameters such as voltage, current, power, irradiance, temperature, and frequency, and collects and summarizes these parameters into the current detection information set to reflect the operating status and environmental conditions of the photovoltaic system at different time points. After preprocessing the current detection information set, it is organized in chronological order to form a time series data set suitable for time series analysis.

[0070] Step S200: Generate first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set.

[0071] In this step, a pre-configured ARIMA (Autoregressive Integrated Moving Average) model is used to model and analyze the time series data set. The ARIMA model generates first prediction information by identifying and capturing the time dependence and trend in the operating parameters of the photovoltaic power generation system, and predicts the operating status of the photovoltaic power generation system at future time points. At the same time, the difference between the actual monitored data and the predicted value of the ARIMA model is calculated to form a residual sequence.

[0072] Step S300: Input the residual sequence into a preset LSTM network to generate second prediction information.

[0073] In this step, the LSTM (Long Short-Term Memory) network, a deep learning model capable of capturing long-term dependencies in time series, is suitable for processing and predicting data with complex temporal characteristics. Taking the residual sequence as the input and passing it to the pre-configured LSTM network, through the analysis of the residual sequence, the LSTM network can identify and learn the non-linear patterns and potential trends that the ARIMA model fails to capture, thereby generating more accurate second prediction information.

[0074] Step S400: Generate preliminary fault information by combining the first prediction information and the second prediction information.

[0075] In this step, comprehensive analysis and fusion are carried out using the first prediction information generated based on the ARIMA model and the second prediction information generated based on the LSTM network. By combining the results of these two prediction methods, the advantages of the ARIMA model in capturing linear trends and time dependencies, as well as the ability of the LSTM network in identifying non-linear patterns and long-term dependencies, can be fully utilized to improve the accuracy and reliability of fault prediction.

[0076] Step S500: Input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

[0077] In this step, using the self-attention mechanism of the Transformer model, key features in the preliminary fault information are deeply extracted, and their context correlation relationships in time and space are analyzed. Through deep feature extraction and context analysis, the model can accurately diagnose the cause of the fault, predict the development trend of the fault, and evaluate its potential impact on the entire photovoltaic power generation system. Based on these analysis results, the Transformer model generates detailed auxiliary decision-making information, including fault type, occurrence time, affected range, emergency handling suggestions, and preventive maintenance measures, etc. The auxiliary decision-making information is sent to the relevant management personnel in a timely manner through appropriate communication channels (such as the management system interface, mobile application notifications, or emails), ensuring that they can quickly and accurately understand the system operation status and take necessary countermeasures, thereby improving the overall operation efficiency and reliability of the photovoltaic power generation system.

[0078] In this embodiment, through comprehensive data collection, preprocessing, multi-model prediction fusion, and deep learning analysis, efficient and accurate fault prediction and decision support are achieved. First, key equipment of the photovoltaic power generation system is monitored in real time and operation parameters are collected, and these data are organized into a time series data set for further analysis. The ARIMA model is used to identify and capture the time dependence and trends in the operation parameters of the photovoltaic power generation system, and then the LSTM network is used to identify and learn the non-linear patterns and potential trends that the ARIMA model fails to capture. By combining the prediction information of the ARIMA and LSTM models, the accuracy and reliability of fault prediction are significantly improved. Then, the Transformer model is used to perform deep feature extraction and context correlation analysis on the preliminary fault detection information, accurately diagnose the cause of the fault, predict the development trend of the fault, and evaluate its potential impact on the entire photovoltaic power generation system. Based on these in-depth analysis results, the system generates detailed auxiliary decision-making information, including the type of fault, the occurrence time, the affected range, emergency handling suggestions, and preventive maintenance measures, etc. These auxiliary decision-making information is timely transmitted to relevant management personnel, enabling the management personnel to accurately understand the fault status of the photovoltaic power generation system, and then dispatching appropriate maintenance personnel to carry corresponding tools and materials to handle the fault, avoiding the problem of low efficiency of fault handling caused by multiple round trips of personnel due to insufficient personnel capabilities and maintenance tools and materials.

[0079] In one embodiment, the step S100: obtaining the current detection information set of the power generation system, and preprocessing the current detection information set to generate a time series data set, includes:

[0080] Step S110: obtaining the current detection information set of the power generation system, and performing data fusion preprocessing on the current detection information set to generate a fusion information set, where the detection information set includes environmental detection information, system operation information, and status detection information.

[0081] In this step, first, various data such as environmental detection information, system operation information, and status detection information of the photovoltaic power generation system are collected. These detection information cover various parameters required for system operation, such as environmental conditions, equipment operation status, etc. The data fusion technology is used to preprocess these multi-source data, eliminate redundancy, improve the integrity and consistency of the data, and generate a comprehensive fusion information set.

[0082] Step S120: using the locally weighted regression interpolation algorithm to perform adaptive data completion on the missing values of the fusion information set, and generating a completed information set, and performing fitting correction on the completed information set through the random forest algorithm, and generating a corrected information set.

[0083] Step S130: Based on the correction information set, perform outlier screening using the z-score method to generate an outlier set. Perform outlier screening on the outlier set using the Isolation Forest algorithm to generate a confirmed outlier set. Based on the locally weighted regression interpolation algorithm, generate a screening information set according to the complement detection information set and the confirmed outlier set.

[0084] In this step, based on the z-score method, by calculating the deviation of data points from the mean, outliers that significantly deviate from the normal range in the correction information set are identified, and these outliers are combined into an outlier set. The Isolation Forest algorithm is used to further confirm the initially screened outlier set to generate a final confirmed outlier set. Combining the complement detection information set and the confirmed outlier set, the locally weighted regression interpolation algorithm is used to generate a screening information set to ensure the accuracy and integrity of the data.

[0085] Step S140: Obtain the time series data of each detection feature in the screening information set, calculate the median and the median absolute deviation value of each feature, and normalize the screening information set according to the absolute deviation value to generate a normalized information set.

[0086] Step S150: Standardize the normalized information set using the z-score method to generate a standard information set, and perform feature engineering processing on the standard information set to generate a feature extraction information set, where the feature extraction information set includes statistical features, periodic features, and frequency domain features.

[0087] In this step, the z-score method is used to standardize the normalized information set to further improve the consistency and comparability of the data. Feature engineering processing is performed on the standard information set to generate a feature extraction information set. Through the extracted statistical features, periodic features, and frequency domain features, important patterns and rules in the data can be captured, providing feature support for subsequent fault prediction and analysis.

[0088] In this embodiment, through data fusion preprocessing, environmental detection information, system operation information, and status detection information are integrated into a comprehensive fused information set to eliminate redundancy and improve the integrity and consistency of the data. Then, the locally weighted regression interpolation algorithm is used to adaptively complete the missing values, and the random forest algorithm is used to fit and correct the completed data to ensure the accuracy of the data. Subsequently, outlier screening is performed based on the z-score method and the isolation forest algorithm to generate a confirmed outlier set, and a screening information set is generated by combining the completed information, thereby removing noise data and enhancing the reliability of the data. Then, by calculating the median and median absolute deviation value of each detection feature, the screening information set is normalized to eliminate the dimensional difference between different features and ensure that the data is compared on the same scale. Finally, the z-score method is used to standardize the normalized information set, and feature engineering processing is performed to extract statistical features, periodic features, and frequency domain features to comprehensively capture important patterns and rules in the data. Through the above preprocessing of the data, the integrity, accuracy, and consistency of the data are improved, the influence of anomalies and noise is reduced, and the feature expression ability is enhanced.

[0089] In one embodiment, step S200: generating first prediction information and a residual sequence based on the time series data set according to a preset ARIMA model includes:

[0090] Step S210: Select target parameters according to a preset historical fault database and expert experience database.

[0091] In this step, according to the operating characteristics and fault modes of the photovoltaic power generation system, as well as the historical fault database and expert experience database, the operating parameters that need to be monitored key are determined. For example, if abnormal voltage and current are likely to cause system failures with a relatively high probability, then the voltage parameter and current parameter are selected as target parameters to improve the pertinence and efficiency of fault detection and concentrate resources on monitoring the parameters that have the greatest impact on the system.

[0092] Step S220: Traverse the time series data set to obtain the operating parameter sequences related to the target parameters, and generate a target operating parameter sequence according to each relevant operating parameter sequence.

[0093] In this step, there may be correlations between different operating parameters. Traversing the entire time series data set to extract the operating parameter sequences related to the selected target parameters and generating a target operating parameter sequence specifically for prediction helps to build a more comprehensive prediction model and improve the accuracy and reliability of the prediction.

[0094] Step S230: Align and combine each of the target operating parameter sequences according to timestamps to generate a target sequence, and input the target sequence into a preset ARIMA model to generate first prediction information.

[0095] Step S240: Calculate the prediction error based on the target sequence and the first prediction information, and generate a residual sequence.

[0096] In this embodiment, the accuracy and reliability of fault detection and prediction are significantly improved through a systematic data processing flow. First, according to the operating characteristics of the photovoltaic system and known fault modes, combined with the historical fault database and the expert experience database, key operating parameters that need to be monitored with emphasis are selected, such as voltage and current, to ensure the pertinence and efficiency of monitoring. Subsequently, traverse the time series data set, extract the operating parameter sequences related to the selected target parameters, and generate the target operating parameter sequences specifically for prediction, making full use of the correlation of multi-dimensional data and enhancing the input information volume of the prediction model. Then, align the target operating parameter sequences according to the timestamps and combine them into a unified target sequence, and input it into the preset ARIMA model to generate the first prediction information, ensuring the consistency of the data and the accuracy of the model prediction. By calculating the error between the actual value of the target sequence and the predicted value of the ARIMA model, a residual sequence is generated. The residual sequence is used to evaluate the prediction performance of the ARIMA model and provide a basis for subsequent fault detection and model optimization. Overall, in this embodiment, by selectively choosing key monitoring parameters, using the multi-parameter correlation to enhance the prediction input, ensuring data consistency, and optimizing the model through residual analysis, the accuracy and efficiency of photovoltaic system fault detection are improved.

[0097] In one embodiment, the step S300: input the residual sequence into a preset LSTM network to generate the second prediction information, includes:

[0098] Step S310: Extract time-dependent features from the residual sequence through a sliding window algorithm, and generate an input feature sequence for the residual sequence based on a set time step.

[0099] In this step, the sliding window method can effectively capture the temporal dependence relationship in the time series data. By sliding the window within the time step to extract continuous data segments, the model can learn the dynamic change pattern and potential temporal dependence of the data.

[0100] Step S320: Extract statistical features, periodic features, and frequency domain features from the input feature sequence, and fuse the time-dependent features, the statistical features, the periodic features, and the frequency domain features to generate an enhanced feature sequence.

[0101] In this step, by extracting statistical, periodic, and frequency domain features, different characteristics of the data are comprehensively described, enhancing the feature expression ability. Fusing multiple features enables the model to understand the data from multiple perspectives, improves the recognition ability of complex patterns, and enhances the accuracy and reliability of non-linear prediction.

[0102] Step S330: Input the enhanced feature sequence into a preset LSTM network for deep modeling to generate first non-linear prediction information.

[0103] Step S340: Perform short-term and medium-term non-linear prediction and trend analysis on the first non-linear prediction information based on a multi-step prediction algorithm to generate second non-linear prediction information.

[0104] Step S350: Filter and correct the second non-linear prediction information based on the expert experience library and generate second prediction information.

[0105] In this step, through the filtering and correction of expert knowledge, false alarms and missed alarms are reduced, ensuring the accuracy and reliability of the prediction information. At the same time, the prediction ability of the machine learning model is combined with expert experience to enhance the professionalism and pertinence of decision-making support.

[0106] In this embodiment, the accuracy and reliability of fault detection and prediction are significantly improved through a systematic data processing process. First, the sliding window algorithm is used to extract time-dependent features from the residual sequence to generate input feature sequences, ensuring that the model can capture the dynamic changes and temporal dependencies of the data. Subsequently, multi-dimensional feature extraction is performed on these input feature sequences to generate statistical features, periodic features, and frequency domain features, and these features are fused with the time-dependent features to form enhanced feature sequences, thereby comprehensively characterizing the complex patterns of the data. Then, the enhanced feature sequences are input into a preset LSTM network for deep modeling to generate a first set of non-linear prediction information, making full use of the LSTM's ability to capture long-term dependencies and non-linear relationships. Based on the multi-step prediction algorithm, trend analysis of the first non-linear prediction information is performed for short-term and medium-term to generate a second set of non-linear prediction information, expanding the prediction time range and early warning ability. Finally, the second non-linear prediction information is filtered and corrected in combination with the expert experience library to generate second prediction information optimized by professional knowledge, enhancing the credibility and practicality of the prediction results. Overall, in this embodiment, through deep feature extraction, advanced prediction models, multi-step prediction, and expert knowledge fusion, not only the prediction accuracy and robustness of the key operating parameters of the photovoltaic system are improved, but also the early warning ability and decision-making support effect of the system are enhanced, reducing false alarms and missed alarms, and ensuring the stable operation of the photovoltaic power generation system in a complex environment.

[0107] In one embodiment, the step S400: Generate preliminary fault information according to the first prediction information and the second prediction information, includes:

[0108] Step S410: Align the timestamps of the first prediction information and the second prediction information to generate an aligned prediction information set.

[0109] In this step, since different models may generate prediction results at different time steps, unaligned data will lead to inaccurate evaluation. Through time alignment, the fairness and accuracy of comparison can be ensured, the errors caused by time bias can be effectively eliminated, and subsequent differential analysis and comprehensive prediction can be made more accurate.

[0110] Step S420: Perform a differential evaluation and analysis of the first prediction information and the second prediction information within the same time interval based on mutual information measurement, and generate a differential analysis index.

[0111] In this step, using mutual information measurement can capture non-linear and complex dependency relationships, deeply understand the differences in the description of the system state by different prediction models within the same time period, and thus provide a basis for the generation of weight parameters.

[0112] Step S430: Generate weight parameters according to the differential analysis index, and generate comprehensive prediction information according to the first prediction information, the second prediction information and the weight parameters.

[0113] In this step, according to the degree of difference between the prediction results, the contribution ratio of each prediction information in the comprehensive result is dynamically adjusted. This embodiment adopts a weighted fusion strategy based on the degree of difference, which improves the flexibility and adaptability of the comprehensive prediction, can dynamically adjust the weight allocation of the prediction results according to real-time data, and thus enhances the accuracy and robustness of the overall prediction.

[0114] Step S440: Generate fault matching information based on the comprehensive prediction information, the historical fault database and the expert experience database according to the pattern matching algorithm, where the fault matching information includes a historical fault identification flag and an expert rule flag.

[0115] In this step, the pattern matching algorithm is used to compare the current prediction result with past fault cases, identify potential fault patterns, and provide targeted handling suggestions based on historical data and expert knowledge, enhancing the depth and breadth of decision support.

[0116] Step S450: Generate preliminary fault information based on the fault matching information according to the naive Bayes algorithm, where the preliminary fault information includes a historical fault matching flag, a fault type, a fault probability and a fault severity.

[0117] In this step, based on the fault matching information, the naive Bayes algorithm is used to generate preliminary fault information, including a historical fault matching flag, a fault type, a fault probability and a fault severity. The naive Bayes algorithm can effectively classify and evaluate the fault state of the current system by calculating the posterior probabilities of different fault types and combining historical matches and expert rules.

[0118] In this embodiment, the accuracy and reliability of fault detection and prediction are significantly improved through a systematic data processing flow. First, the timestamps of the first prediction information and the second prediction information are aligned to generate an aligned prediction information set, ensuring the temporal consistency and comparability of the two sets of prediction data. Subsequently, based on mutual information measurement, a differential evaluation and analysis of the two sets of prediction information within the same time interval is performed to generate differential analysis metrics, effectively quantifying the correlation and degree of difference between the prediction results. Then, weight parameters are generated according to the differential analysis metrics, and the first prediction information and the second prediction information are weighted and fused using the weight parameters to generate comprehensive prediction information, thereby improving the accuracy and robustness of the prediction results. Then, a pattern matching algorithm is used to match the comprehensive prediction information with the historical fault database and the expert experience database to generate fault matching information containing historical fault identification marks and expert rule marks, ensuring the comprehensiveness and professionalism of fault identification. Finally, based on the fault matching information, a naive Bayes algorithm is used to generate preliminary fault information, including historical fault matching marks, fault types, fault probabilities, and fault severity levels, providing detailed and reliable fault diagnosis results. Through time alignment, differential evaluation, weighted fusion, pattern matching, and probability analysis, this embodiment comprehensively improves the accuracy of fault detection in the photovoltaic power generation system and the effectiveness of decision-making support.

[0119] In one embodiment, step S500: inputting the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis and generating auxiliary decision-making information, and sending the auxiliary decision-making information to the management personnel, includes:

[0120] Step S510: retrieving the corresponding fault case text information from the historical fault database according to the historical fault matching mark, querying the corresponding processing suggestion text from the expert experience database according to the expert rule mark, and performing vector representation on the fault case text information and the processing suggestion text based on a text encoding algorithm to generate context embedding data.

[0121] In this step, historical fault cases and expert suggestions are converted into a numerical form that can be processed by a machine learning model, thereby providing rich semantic information for subsequent deep feature extraction and decision analysis. By combining historical data and expert knowledge, the system can more comprehensively understand fault patterns, improve the accuracy of fault identification and the relevance of decision-making suggestions, and enhance the overall fault detection and response capabilities.

[0122] Step S520: performing feature splicing and weighted integration according to the preliminary fault detection information and the context embedding data to generate a fused feature representation.

[0123] In this step, the real-time operating status of the photovoltaic power generation system is combined with historical fault cases and expert suggestions to form a comprehensive feature set for subsequent analysis by the deep learning model. Through feature splicing and weighted integration, multi-source information can be considered simultaneously, improving the richness and diversity of feature representation, thereby enhancing the model's ability to recognize complex fault patterns.

[0124] Step S530: Input the fused feature representation into a pre-trained Transformer model, perform deep feature extraction and context correlation analysis on the fused feature through the multi-head attention mechanism, and generate deep correlation features.

[0125] Step S540: Generate auxiliary decision-making candidate information based on the deep correlation features and the expert experience library.

[0126] In this step, the complex features extracted by the deep learning model are combined with expert knowledge, and through the policy inference mechanism, specific fault handling suggestions and maintenance strategies are provided. Using the expert experience library, the system can incorporate professional judgments on a data-driven basis to generate more targeted and practical decision-making information.

[0127] Step S550: Format and preferentially screen the auxiliary decision-making candidate information to generate auxiliary decision-making information, where the auxiliary decision-making information includes executable maintenance strategies, suggestions on maintenance timing, and resource allocation plans.

[0128] In this step, the initially generated decision-making information is organized into an easy-to-understand and implement form to ensure that managers can quickly obtain and apply these suggestions. Through formatting and screening, the system can eliminate redundant or irrelevant information, highlight key decision-making points, and improve the effectiveness and operability of decision-making support.

[0129] Step S560: Send the auxiliary decision-making information to the manager and obtain feedback information, and store the feedback information in the historical fault database and the expert experience library.

[0130] In this step, the generated auxiliary decision-making information is sent to the manager and feedback information is obtained. At the same time, this feedback information is stored in the historical fault database and the expert experience library to achieve dynamic update of knowledge and continuous optimization of the system. Establish a closed-loop feedback mechanism. Through the actual operation feedback of managers, continuously enrich and improve the system's fault database and expert experience library. By storing and analyzing the feedback information, the system can learn and adapt to new fault patterns and handling methods, improving the accuracy and effectiveness of future fault detection and decision-making support.

[0131] In this embodiment, by integrating historical fault cases and expert experience, the accuracy and reliability of fault detection and decision-making are significantly improved. First, the system retrieves and vectorizes fault cases and processing suggestion texts from the historical fault database and the expert experience database according to the historical fault matching identifier and the expert rule identifier, and generates context embedding data. Subsequently, the preliminary fault detection information is subjected to feature splicing and weighted integration with these context embedding data to form a fused feature representation, which is input into a pre-trained Transformer model. The features are deeply extracted through the multi-head attention mechanism to generate deeply correlated features. Based on these deeply correlated features, the system combines the expert experience database for policy inference, generates auxiliary decision-making candidate information, and through result formatting and optimal screening, finally provides auxiliary decision-making information including executable maintenance strategies, overhaul timing suggestions, and resource allocation plans. Finally, the system sends this decision-making information to the management personnel and collects feedback information to continuously update the historical fault database and the expert experience database to achieve dynamic optimization of knowledge.

[0132] In one embodiment, step S430: generating a weight parameter according to the difference analysis index, and generating comprehensive prediction information according to the first prediction information, the second prediction information, and the weight parameter, includes:

[0133] Step S431: generating a weight parameter according to the difference analysis index based on the following formula:

[0134]

[0135] where ω(t) is the weight parameter, t is the ordinal number of the time step, α is a preset change sensitivity hyperparameter, β is a preset expected difference threshold, and D(t) is the difference analysis index.

[0136] Step S432: generating a basic fusion result for the first prediction information and the second prediction information based on the following formula according to the weight parameter:

[0137] P′(t) = ω(t)μ1(t) + [1 - ω(t)]μ2(t);

[0138] where P′(t) is the basic fusion result, t is the ordinal number of the time step, ω(t) is the weight parameter, μ1(t) is the first prediction information, and μ2(t) is the second prediction information.

[0139] Step S433: generating a relative difference degree according to the first prediction information and the second prediction information based on the following formula:

[0140]

[0141] Wherein, R(t) is the relative difference degree, t is the ordinal number of the time step, μ1(t) is the first prediction information, μ2(t) is the second prediction information, and γ is a preset positive constant.

[0142] Step S434: Correct the basic fusion result based on the relative difference degree according to the following formula to generate a comprehensive prediction result:

[0143] P(t) = P′(t) - η·R(t)·[P′(t) - ξ];

[0144] Wherein, P(t) is the comprehensive prediction result, t is the ordinal number of the time step, P′(t) is the basic fusion result, η is a preset correction amplitude adjustment parameter, R(t) is the relative difference degree, and ξ is a preset historical mean value.

[0145] In this embodiment, the contribution ratio of the two prediction information is flexibly adjusted by using the weight parameter. At the same time, the deficiency of the model at a specific time step is compensated by the difference degree correction. Combining the historical mean value and the correction amplitude adjustment mechanism ensures the accuracy and stability of the prediction result. The advantage of the scheme is strong dynamic adaptability. It can adjust the fusion strategy according to the real-time difference index, give full play to the complementarity of multi-source information, and significantly improve the prediction accuracy. In addition, by introducing the normalization and smoothing factor, the robustness of the model to noise and extreme values is effectively enhanced, enabling it to adapt to the changing environment of the photovoltaic power generation system.

[0146] In one embodiment, the ARIMA model is established based on the following steps:

[0147] Step S610: Obtain the conventional detection information set of the power generation system, generate a conventional time series data set according to the conventional detection information set, and perform multi-granularity partitioning on the conventional time series data set to generate a modeling sequence and a verification sequence.

[0148] In this step, first, various types of conventional detection information of the power generation system are collected. These information include environmental data (such as temperature, humidity, light intensity, etc.), system operation data (such as voltage, current, power, etc.), and equipment status data (such as equipment operation status, fault information, etc.). These data are usually time series data. By preprocessing these data, they are converted into a standard time series data set. To ensure the generalization ability and robustness of the model, these time series data sets will be further divided into a modeling sequence and a verification sequence. This multi-granularity partitioning method can model and verify the data through different time scales (such as hours, days, weeks, etc.), helping the system capture different features from short-term changes to long-term trends.

[0149] Step S620: Decompose the modeling sequence through variational mode decomposition to generate multiple intrinsic mode components, establish corresponding initial ARIMA sub-models according to each of the intrinsic mode components, and iteratively search for and update the parameters of the initial ARIMA sub-models through the Bayesian optimization algorithm to generate ARIMA sub-models.

[0150] Step S630: Perform rolling prediction on the validation sequence based on the ARIMA sub-models using a fixed-length sliding window algorithm to generate a first prediction result.

[0151] In this step, a fixed-length sliding window algorithm is used to perform rolling prediction on the validation sequence. Each time, a fixed-sized time window (such as historical data of the past few days or hours) is used to predict the system state for a period of time in the future. This enables the model to be updated based on the latest data at each time step, avoiding the influence of outdated data and maintaining the timeliness of the prediction.

[0152] Step S640: Compare the first prediction result with the true value of the validation sequence to generate a residual sequence through residual refinement, and optimize the ARIMA sub-model according to the residual sequence to generate an optimized ARIMA sub-model.

[0153] In this step, through residual refinement, possible biases or error patterns in the model can be identified, and then the ARIMA sub-model can be optimized according to these residual sequences, thereby improving the fitting accuracy of the model and reducing prediction errors.

[0154] Step S650: Quantify the performance of the optimized ARIMA sub-model according to the residual sequence and generate an error value, and the error value is obtained based on the following formula:

[0155]

[0156] where E is the error value, N is the number of time steps in the validation sequence, t is the ordinal number of the time step, c t is the weight factor corresponding to the time step t, y t is the true value of the validation sequence at time step t, y' t is the model prediction value of the validation sequence at time step t, and ε is a preset positive constant.

[0157] Step S670: Generate sub-model weight parameters according to the error value, the historical fault database, and the expert experience database, and construct the ARIMA model according to the ARIMA sub-model and the sub-model weight parameters.

[0158] In this embodiment, through systematic data partitioning, variational mode decomposition, and Bayesian optimization, the multi-level linear trends and dynamic change patterns of system operating parameters are accurately captured and modeled. First, the conventional detection data is segmented into modeling and validation sequences to ensure the accuracy and generalization ability of model training. Then, the complex time series is decomposed into multiple intrinsic mode components through variational mode decomposition, and the parameters of each ARIMA sub-model are iteratively optimized using Bayesian optimization to improve the prediction accuracy. Next, a fixed-length sliding window is used to perform rolling prediction on the validation sequence to generate preliminary prediction results, and the model is further optimized through residual refinement to reduce prediction errors. Finally, the model performance is quantified through an error formula to ensure the reliability of the prediction results. This embodiment combines error quantification with historical fault databases and expert experience databases to generate a weighted-optimized ARIMA model, realizing the reasonable allocation of multi-model weights, improving the prediction accuracy of key parameters of the photovoltaic system and the robustness of the model, and enhancing the sensitivity and response ability to anomalies and faults through continuous optimization and historical data fusion.

[0159] In one embodiment, as Figure 2 shown, a fault-assisted decision-making generation system for a power generation system based on a large model is provided, including a time series data generation module, a first prediction information generation module, a second prediction information generation module, a preliminary fault information generation module, and an auxiliary decision-making information generation module.

[0160] The time series data generation module is used to obtain the current detection information set of the power generation system and preprocess the current detection information set to generate a time series data set. The first prediction information generation module is used to generate first prediction information and a residual sequence based on the time series data set according to a preset ARIMA model. The second prediction information generation module is used to input the residual sequence into a preset LSTM network to generate second prediction information. The preliminary fault information generation module is used to generate preliminary fault information by combining the first prediction information and the second prediction information. The auxiliary decision-making information generation module is used to input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

[0161] In one embodiment, the time series data generation module is further used for:

[0162] Obtain the current detection information set of the power generation system, and perform data fusion preprocessing on the current detection information set to generate a fusion information set, where the detection information set includes environmental detection information, system operation information, and status detection information; use the locally weighted regression interpolation algorithm to perform adaptive data completion on the missing values of the fusion information set, and generate a completed information set, and perform fitting correction on the completed information set through the random forest algorithm to generate a corrected information set; perform anomaly screening on the corrected information set based on the z-score method to generate an outlier set, perform anomaly screening on the outlier set through the isolation forest algorithm to generate a confirmed outlier set, and generate a screening information set based on the locally weighted regression interpolation algorithm according to the completed detection information set and the confirmed outlier set; obtain the time series data of each detection feature in the screening information set, calculate the median and the median absolute deviation value of each feature, and perform normalization processing on the screening information set according to the absolute deviation value to generate a normalized information set; perform standardization on the normalized information set based on the z-score method to generate a standard information set, and perform feature engineering processing on the standard information set to generate a feature extraction information set, where the feature extraction information set includes statistical features, periodic features, and frequency domain features.

[0163] In one embodiment, the first prediction information generation module is further configured to:

[0164] Select target parameters according to a preset historical fault database and expert experience database; traverse the time series data set to obtain the operation parameter sequences related to the target parameters, and generate a target operation parameter sequence according to each relevant operation parameter sequence; align and combine each target operation parameter sequence according to the time stamp to generate a target sequence, input the target sequence into a preset ARIMA model to generate first prediction information; calculate the prediction error according to the target sequence and the first prediction information, and generate a residual sequence.

[0165] In one embodiment, the first prediction information generation module is further configured to:

[0166] Obtain the conventional detection information set of the power generation system, generate a conventional time series data set according to the conventional detection information set, and perform multi-granularity partitioning on the conventional time series data set to generate a modeling sequence and a verification sequence; decompose the modeling sequence through variational mode decomposition to generate multiple intrinsic mode components, establish corresponding initial ARIMA sub-models according to each intrinsic mode component, and iteratively search and update the parameters of the initial ARIMA sub-models through the Bayesian optimization algorithm to generate ARIMA sub-models; based on the ARIMA sub-models, use the fixed-length sliding window algorithm to perform rolling prediction on the verification sequence to generate a first prediction result; compare the first prediction result with the true value of the verification sequence to generate a residual sequence through residual refinement, and optimize the ARIMA sub-models according to the residual sequence to generate optimized ARIMA sub-models; perform performance quantification on the optimized ARIMA sub-models according to the residual sequence and generate an error value, and the error value is obtained based on the following formula:

[0167]

[0168] where E is the error value, N is the number of time steps of the verification sequence, t is the ordinal number of the time step, c t is the weight factor corresponding to the time step t, y t is the true value of the verification sequence at the time step t, y′ t is the model prediction value of the verification sequence at the time step t, and ε is a preset positive constant; generate sub-model weight parameters according to the error value, the historical fault database and the expert experience database, and construct the ARIMA model according to the ARIMA sub-models and the sub-model weight parameters.

[0169] In one embodiment, the second prediction information generation module is further configured to:

[0170] Extract time-dependent features from the residual sequence through a sliding window algorithm, and generate an input feature sequence for the residual sequence based on a set time step length; perform feature extraction on the input feature sequence to generate statistical features, periodic features and frequency domain features, and fuse the time-dependent features, the statistical features, the periodic features and the frequency domain features to generate an enhanced feature sequence; input the enhanced feature sequence into a preset LSTM network for deep modeling to generate first non-linear prediction information; perform short-term and medium-term non-linear prediction and trend analysis on the first non-linear prediction information based on a multi-step prediction algorithm to generate second non-linear prediction information; filter and correct the second non-linear prediction information based on the expert experience library, and generate second prediction information.

[0171] In one embodiment, the preliminary fault information generation module is further configured to:

[0172] Align the timestamps of the first prediction information and the second prediction information to generate an aligned prediction information set; perform a difference evaluation analysis of the first prediction information and the second prediction information within the same time interval based on mutual information measurement, and generate a difference analysis index; generate a weight parameter according to the difference analysis index, and generate comprehensive prediction information according to the first prediction information, the second prediction information, and the weight parameter; generate a fault matching information based on the comprehensive prediction information, the historical fault database, and the expert experience database using a pattern matching algorithm, where the fault matching information includes a historical fault identification flag and an expert rule flag; generate preliminary fault information based on the fault matching information using a Naive Bayes algorithm, where the preliminary fault information includes a historical fault matching flag, a fault type, a fault probability, and a fault severity level.

[0173] In one embodiment, the preliminary fault information generation module is further configured to: generate a weight parameter according to the difference analysis index based on the following formula:

[0174]

[0175] where ω(t) is the weight parameter, t is the ordinal number of the time step, α is a preset change sensitivity hyperparameter, β is a preset expected difference threshold, and D(t) is the difference analysis index;

[0176] Generate a basic fusion result for the first prediction information and the second prediction information based on the following formula according to the weight parameter:

[0177] P′(t) = ω(t)μ1(t) + [1 - ω(t)]μ2(t);

[0178] where P′(t) is the basic fusion result, t is the ordinal number of the time step, ω(t) is the weight parameter, μ1(t) is the first prediction information, and μ2(t) is the second prediction information;

[0179] Generate a relative difference degree for the first prediction information and the second prediction information based on the following formula:

[0180]

[0181] where R(t) is the relative difference degree, t is the ordinal number of the time step, μ1(t) is the first prediction information, μ2(t) is the second prediction information, and γ is a preset positive constant;

[0182] Correct the basic fusion result based on the relative difference degree according to the following formula to generate a comprehensive prediction result:

[0183] P(t) = P'(t) - η·R(t)·[P'(t) - ξ];

[0184] Wherein, P(t) is the comprehensive prediction result, t is the ordinal number of the time step, P'(t) is the basic fusion result, η is a preset correction amplitude adjustment parameter, R(t) is the relative difference degree, and ξ is the preset historical mean value.

[0185] In one embodiment, the auxiliary decision information generation module is further configured to:

[0186] Retrieve the corresponding fault case text information from the historical fault database according to the historical fault matching identifier, query the corresponding processing suggestion text from the expert experience database according to the expert rule identifier, perform vectorization representation on the fault case text information and the processing suggestion text based on the text encoding algorithm to generate context embedding data; perform feature splicing and weighted integration on the preliminary fault detection information and the context embedding data to generate a fusion feature representation; input the fusion feature representation into a pre-trained Transformer model, perform deep feature extraction and context correlation analysis on the fusion feature through the multi-head attention mechanism, and generate deep correlation features; perform policy inference on the deep correlation features and the expert experience database to generate auxiliary decision candidate information; perform result formatting and optimal screening on the auxiliary decision candidate information to generate auxiliary decision information, where the auxiliary decision information includes executable maintenance strategies, maintenance timing suggestions, and resource allocation plans; send the auxiliary decision information to the management personnel and obtain feedback information, and store the feedback information in the historical fault database and the expert experience database.

[0187] In one embodiment, a computer device is provided, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for generating auxiliary decision for power generation system faults based on a large model are implemented.

[0188] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned method for generating auxiliary decision for power generation system faults based on a large model are implemented.

[0189] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0190] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0191] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for generating auxiliary decision-making on power generation system faults based on large models, characterized in that, Including: Obtain the current detection information set of the power generation system, and preprocess the current detection information set to generate a time series data set; Generate first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set; Input the residual sequence into a preset LSTM network to generate second prediction information; Generate preliminary fault information by combining the first prediction information and the second prediction information; Input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

2. The method for generating a fault auxiliary decision of a power generation system based on a large model according to claim 1, wherein Generating first prediction information and a residual sequence based on the preset ARIMA model according to the time series data set includes: Select target parameters according to the preset historical fault database and expert experience database; Traverse the time series data set to obtain the operation parameter sequences related to the target parameters, and generate the target operation parameter sequence according to each related operation parameter sequence; Align and combine each target operation parameter sequence according to the time stamp to generate a target sequence, and input the target sequence into a preset ARIMA model to generate first prediction information; Calculate the prediction error according to the target sequence and the first prediction information, and generate a residual sequence.

3. The method for generating a fault auxiliary decision of a power generation system based on a large model according to claim 2, wherein Inputting the residual sequence into a preset LSTM network to generate second prediction information includes: Extract time-dependent features from the residual sequence through a sliding window algorithm, and generate an input feature sequence based on the set time step for the residual sequence; Extract features from the input feature sequence to generate statistical features, periodic features and frequency domain features, and fuse the time-dependent features, the statistical features, the periodic features and the frequency domain features to generate an enhanced feature sequence; Input the enhanced feature sequence into a preset LSTM network for deep modeling to generate first non-linear prediction information; Perform short-term and medium-term non-linear prediction and trend analysis on the first non-linear prediction information based on a multi-step prediction algorithm to generate second non-linear prediction information; Filter and correct the second non-linear prediction information based on the expert experience database, and generate second prediction information.

4. The method for generating a fault auxiliary decision of a power generation system based on a large model according to claim 3, wherein Generating preliminary fault information by combining the first prediction information and the second prediction information includes: Align the time stamps of the first prediction information and the second prediction information to generate an aligned prediction information set; Perform a difference evaluation analysis on the first prediction information and the second prediction information within the same time interval based on mutual information measurement, and generate a difference analysis index; Generate a weight parameter according to the difference analysis index, and generate comprehensive prediction information according to the first prediction information, the second prediction information and the weight parameter; Generate fault matching information based on the comprehensive prediction information, the historical fault database and the expert experience database based on a pattern matching algorithm, where the fault matching information includes a historical fault identification flag and an expert rule flag; Generate preliminary fault information based on the fault matching information using the Naive Bayes algorithm, where the preliminary fault information includes a historical fault matching identifier, a fault type, a fault probability, and a fault severity level.

5. The method for generating a fault auxiliary decision of a power generation system based on a large model according to claim 4, wherein Input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information. Send the auxiliary decision-making information to the management staff, including: Retrieve the corresponding fault case text information from the historical fault database according to the historical fault matching identifier, query the expert experience database according to the expert rule identifier to obtain the corresponding processing suggestion text, and perform vectorization representation on the fault case text information and the processing suggestion text based on the text encoding algorithm to generate context embedding data; Perform feature splicing and weighted integration based on the preliminary fault detection information and the context embedding data to generate a fused feature representation; Input the fused feature representation into a pre-trained Transformer model, perform deep feature extraction and context correlation analysis on the fused feature through the multi-head attention mechanism, and generate a deep correlation feature; Generate auxiliary decision-making candidate information based on the deep correlation feature and the expert experience database; Perform result formatting and optimal screening on the auxiliary decision-making candidate information to generate auxiliary decision-making information, where the auxiliary decision-making information includes executable maintenance strategies, inspection time suggestions, and resource allocation plans; Send the auxiliary decision-making information to the management staff and obtain feedback information, and store the feedback information in the historical fault database and the expert experience database.

6. The method for generating a fault-assisted decision of a power generation system based on a large model according to any one of claims 2-5, wherein The ARIMA model is established based on the following steps: Obtain the conventional detection information set of the power generation system, generate a conventional time series data set according to the conventional detection information set, and perform multi-granularity partitioning on the conventional time series data set to generate a modeling sequence and a validation sequence; Decompose the modeling sequence through variational mode decomposition to generate multiple intrinsic mode components, establish a corresponding initial ARIMA sub-model according to each intrinsic mode component, and perform iterative search and update on the parameters of the initial ARIMA sub-model through the Bayesian optimization algorithm to generate an ARIMA sub-model; Perform rolling prediction on the validation sequence based on the ARIMA sub-model using the fixed-length sliding window algorithm to generate a first prediction result; Compare the first prediction result with the true value of the validation sequence to refine the residuals and generate a residual sequence, and optimize the ARIMA sub-model according to the residual sequence to generate an optimized ARIMA sub-model; Quantify the performance of the optimized ARIMA sub-model according to the residual sequence and generate an error value, where the error value is obtained based on the following formula: Where E is the error value, N is the number of verification sequence time steps, t is the ordinal number of the time step, c t is the weight factor corresponding to the time step t, y t is the true value of the verification sequence at the time step t, y′ t is the model prediction value of the verification sequence at the time step t, and ε is a preset positive constant; Generate sub-model weight parameters according to the error value, the historical fault database, and the expert experience database, and construct the ARIMA model according to the ARIMA sub-model and the sub-model weight parameters.

7. The method for generating a fault auxiliary decision of a power generation system based on a large model according to any one of claims 1-5, characterized in that Obtain the current detection information set of the power generation system, and preprocess the current detection information set to generate a time series data set, including: Obtain the current detection information set of the power generation system, and perform data fusion preprocessing on the current detection information set to generate a fusion information set, where the detection information set includes environmental detection information, system operation information, and status detection information; Adopt the locally weighted regression interpolation algorithm to adaptively fill in the missing values for the fusion information set, and generate a completed information set. Fit and correct the completed information set through the random forest algorithm, and generate a corrected information set; Based on the z-score method, perform anomaly screening on the corrected information set to generate an outlier set, perform anomaly screening on the outlier set through the isolation forest algorithm to generate a confirmed outlier set, and generate a screening information set based on the locally weighted regression interpolation algorithm according to the completed detection information set and the confirmed outlier set; Obtain the time series data of each detection feature in the screening information set, calculate the median and the median absolute deviation value of each feature, and perform normalization processing on the screening information set according to the absolute deviation value to generate a normalized information set; Standardize the normalized information set based on the z-score method to generate a standard information set, and perform feature engineering processing on the standard information set to generate a feature extraction information set, where the feature extraction information set includes statistical features, periodic features, and frequency domain features.

8. A fault-assisted decision-making generation system for a power generation system based on a large model, characterized in that, Including: A time series data generation module, configured to obtain the current detection information set of the power generation system, and preprocess the current detection information set to generate a time series data set; A first prediction information generation module, configured to generate first prediction information and a residual sequence based on the time series data set according to a preset ARIMA model; A second prediction information generation module, configured to input the residual sequence into a preset LSTM network to generate second prediction information; A preliminary fault information generation module, configured to generate preliminary fault information by combining the first prediction information and the second prediction information; An auxiliary decision-making information generation module, configured to input the preliminary fault detection information into a pre-trained Transformer model for deep feature extraction and context correlation analysis, and generate auxiliary decision-making information, and send the auxiliary decision-making information to the management personnel.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images