A Method and System for Power System Simulation Curve Analysis Based on Large Language Model

Abstract

This application proposes a method and system for power system simulation curve analysis based on a large language model, comprising: acquiring simulation data and manual analysis results text for various power system operating scenarios; training the simulation data using a layer-by-layer training method based on a deep autoencoder to obtain encoders and decoders corresponding to different stability labels; obtaining the output deviation of each category of samples through the encoder-decoder based on the simulation data corresponding to different labels, the encoder, and the decoder; fine-tuning the large model based on the simulation data and the manual analysis results text to obtain a fine-tuned curve analysis model; acquiring the simulation curve data of the power system to be analyzed, and obtaining the analysis results based on the simulation curve data, the encoder and decoder, the output deviation of each category of samples through the encoder-decoder, and the curve analysis model. This invention, employing the above scheme, can provide decision-making basis in text form while ensuring a certain degree of accuracy in the analysis results.

Description

Technical Field

[0001] This application relates to the field of power system security analysis technology, and in particular to a method and system for power system simulation curve analysis based on a large language model. Background Technology

[0002] Transient stability simulation curve analysis is an important function in power system analysis and control. However, traditional simulation curve analysis relies on manual curve reading, which is time-consuming and labor-intensive.

[0003] Patent CN115908620A proposes a method for displaying transient stability simulation curves, facilitating manual interpretation of the curves, but it does not address how to analyze the simulation curves. Patent CN113488992A proposes a method for determining stability under large disturbances, capable of analyzing simulation curves, but this method only provides the category result of the simulation analysis. In practical applications, it is often desirable to provide the reasons behind the curve analysis results and to promptly alert for potentially erroneous analysis results requiring manual intervention. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, the first objective of this application is to propose a power system simulation curve analysis method based on a large language model, so as to at least solve the technical problem in the prior art that it is difficult to provide text analysis results and error prompts for simulation curve analysis.

[0006] The second objective of this application is to propose a power system simulation curve analysis system based on a large language model.

[0007] To achieve the above objectives, the first aspect of this application proposes a method for analyzing power system simulation curves based on a large language model, comprising:

[0008] Simulation data under various operating scenarios of the power system is acquired, and the simulation data is analyzed manually to obtain the manual analysis result text, which includes the stability label of the simulation data and the reason for the judgment.

[0009] A layer-by-layer training method based on deep autoencoders is used to train on simulation data to obtain encoders and decoders corresponding to different stability labels.

[0010] Based on the simulation data, encoder and decoder corresponding to different stability labels, the maximum deviation of the sample output by the encoder and decoder for each type of stability label is obtained.

[0011] Simulation data from various operating scenarios of the power system are used as input data, and the text of manual analysis results is used as output data. The large model is then fine-tuned to obtain a fine-tuned curve analysis model.

[0012] Input the simulated curve data of the power system to be analyzed into the fine-tuned curve analysis model, and output the analysis result text;

[0013] Based on the analysis results text, the stability label of the simulation curve data is determined, and the simulation curve data is input into the encoder and decoder corresponding to its stability label to obtain the output vector. It is then determined that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, thus determining the analysis result of the power system simulation curve data to be analyzed.

[0014] Optionally, in one embodiment of this application, obtaining simulation data under various operating scenarios of the power system includes:

[0015] Based on the historical operation of the power system and load forecast, K operating scenarios are randomly set, and the node voltage amplitude curves, generator rotor angle curves and generator frequency curves under the K operating scenarios are obtained by power flow calculation and transient simulation calculation.

[0016] The voltage amplitude data of the top S nodes with the largest and smallest average voltage amplitudes are selected sequentially from the node voltage amplitude curves to form the key node voltage amplitude data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0017] The rotor angle data of the top S generators with the largest and smallest average rotor angles are selected sequentially from the generator rotor angle curves to form the key generator rotor angle data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0018] The frequency data of the top S generators with the highest and lowest average frequencies are selected sequentially from the generator frequency curves to form the key generator frequency data. Where n = 1, 2, ..., 2S, and t is the t-th sampling time in the simulation curve, t = 1, ..., T.

[0019] Optionally, in one embodiment of this application, the simulation data is analyzed manually to obtain a text of the manual analysis results, including:

[0020] Based on the safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability labels and judgment reasons as the manual analysis result text. Among them, there are a types of stability labels in the manual analysis result text.

[0021] Optionally, in one embodiment of this application, before using simulation data for training, the method further includes: preprocessing the simulation data, the preprocessing process including:

[0022] Sequentially analyze the key generator rotor angle data under the k-th operating scenario. Standardization was performed to obtain standardized key generator rotor angle data. Represented as:

[0023]

[0024] The key generator frequency data under the k-th operating scenario are analyzed sequentially. Standardization was performed to obtain standardized key generator frequency data. Represented as:

[0025]

[0026] Among them, K s denoted as the number of running scenarios corresponding to the s-th stability label.

[0027] Optionally, in one embodiment of this application, a layer-by-layer training method based on a deep autoencoder, using simulation data for training, yields encoders and decoders corresponding to different stability labels, including:

[0028] The preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data for the scenario corresponding to the s-th stability label are sequentially selected from the simulation data to constitute the input feature data for the scenario corresponding to the s-th stability label. X is the input feature of the j-th running scenario with the s-th stability label, where the j-th running scenario with the s-th stability label corresponds to the j_s-th running scenario among the K running scenarios. j_s Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data

[0029] Define the encoder structure, where the encoder input data is X. j_s The output is a dimension w s eigenvector h sThe encoder's output dimension and hidden layer settings are determined based on actual data and human experience;

[0030] Define the structure of the decoder, where the input data of the decoder is the encoder output of dimension w. s eigenvector h s The output is the dimension and X. j_s Same feature vector The hidden layer settings of the decoder are determined based on the characteristics of the actual data and human experience;

[0031] Based on the input data of all operating scenarios using the s-th stability label and the defined encoder and decoder, the encoder B corresponding to the s-th category label is trained using a layer-by-layer training method of a deep autoencoder and the constructed loss function loss. s and decoder C s The expression for the loss function is:

[0032]

[0033] Where D is the set of running scenario IDs participating in the training, d is the number of running scenarios in set D, and r is the r-th running scenario in set D.

[0034]

[0035] Where ε is the set threshold for the deviation between the predicted value and the true value of the feature.

[0036] Optionally, in one embodiment of this application, based on simulation data, encoders, and decoders corresponding to different stability labels, the maximum deviation output by the encoder and decoder for each type of stability label is obtained, including:

[0037] Will Input to the corresponding encoder B s and decoder C s Then, the output vector is obtained.

[0038] The maximum deviation of all scene samples corresponding to the s-th stability label after passing through the encoder-decoder output is calculated using the following formula:

[0039]

[0040] in,

[0041]

[0042] Optionally, in one embodiment of this application, simulation data from various operating scenarios of the power system are used as input data, and the text of manually analyzed results is used as output data to fine-tune the large model, resulting in a finely tuned curve analysis model, including:

[0043] Choose a large language model with input word vector dimension o;

[0044] The preprocessed simulation data is arranged into three-dimensional data according to the variable number dimension, time dimension, and variable category dimension. The dimension of this three-dimensional data is (2×S)×T×3.

[0045] Construct an input processing layer, which is used to reduce the dimensionality of the three-dimensional data to an o-dimensional vector, and the parameter to be learned in the input processing layer is W;

[0046] The preprocessed simulation data is used as input feature data and passed sequentially through the input processing layer and the large language model to obtain the output results. Based on the output results and the text of the manual analysis results, a loss function is constructed to fine-tune the learning parameters W of the input processing layer and the parameters of the large language model to obtain the fine-tuned curve analysis model.

[0047] Optionally, in one embodiment of this application, determining the analysis result of the power system simulation curve data to be analyzed by judging that the deviation between the input and output is less than the maximum deviation corresponding to its stability label includes:

[0048] Calculate the output vector of the decoder With input feature data (x1,...,x e ,...,x 6×S×T The deviation g of ) is expressed as:

[0049]

[0050] in,

[0051]

[0052] Compare the deviation g with the maximum deviation g corresponding to the z-th stability label. z The size of g, if g ≤ g z If the accuracy of the analysis result text is determined, the analysis result text will be used as the analysis result of the power system simulation curve data to be analyzed.

[0053] Optionally, in one embodiment of this application, the method further includes:

[0054] If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, the analysis result text is judged to be inaccurate. The analysis result is obtained by manually analyzing the power system simulation curve data to be analyzed.

[0055] To achieve the above objectives, a second aspect of the present invention proposes a power system simulation curve analysis system based on a large language model, comprising:

[0056] The data acquisition module is used to acquire simulation data under various operating scenarios of the power system, and to obtain the manual analysis result text through manual analysis of the simulation data. The manual analysis result text includes the stability label of the simulation data and the reason for the judgment.

[0057] The encoder-decoder training module is used for training based on a layer-by-layer training method of deep autoencoders, using simulation data to obtain encoders and decoders corresponding to different stability labels.

[0058] The output deviation statistics module is used to obtain the maximum deviation of the sample corresponding to each stability label through the encoder and decoder based on the simulation data, encoder and decoder corresponding to different stability labels.

[0059] The curve analysis model training module is used to fine-tune the large model by taking simulation data from various operating scenarios of the power system as input data and text of manual analysis results as output data, and to obtain a fine-tuned curve analysis model.

[0060] The curve analysis module is used to input the simulated curve data of the power system to be analyzed into the finely tuned curve analysis model and output the analysis results text.

[0061] The curve analysis module is also used to determine the stability label of the simulation curve data based on the analysis result text, and input the simulation curve data into the encoder and decoder corresponding to its stability label to obtain the output vector. It also determines that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, and determines the analysis result of the power system simulation curve data to be analyzed.

[0062] Optionally, in one embodiment of this application, obtaining simulation data under various operating scenarios of the power system includes:

[0063] Based on the historical operation of the power system and load forecast, K operating scenarios are randomly set, and the node voltage amplitude curves, generator rotor angle curves and generator frequency curves under the K operating scenarios are obtained by power flow calculation and transient simulation calculation.

[0064] The voltage amplitude data of the top S nodes with the largest and smallest average voltage amplitudes are selected sequentially from the node voltage amplitude curves to form the key node voltage amplitude data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0065] The rotor angle data of the top S generators with the largest and smallest average rotor angles are selected sequentially from the generator rotor angle curves to form the key generator rotor angle data. Where n = 1, 2, ..., 2s, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0066] The frequency data of the top S generators with the highest and lowest average frequencies are selected sequentially from the generator frequency curves to form the key generator frequency data. Where n = 1, 2, ..., 2S, and t is the t-th sampling time in the simulation curve, t = 1, ..., T.

[0067] Optionally, in one embodiment of this application, the simulation data is analyzed manually to obtain a text of the manual analysis results, including:

[0068] Based on the safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability labels and judgment reasons as the manual analysis result text. Among them, there are a types of stability labels in the manual analysis result text.

[0069] Optionally, in one embodiment of this application, the data acquisition module is further configured to preprocess the simulation data before training with the simulation data, the preprocessing process including:

[0070] Sequentially analyze the key generator rotor angle data under the k-th operating scenario. Standardization was performed to obtain standardized key generator rotor angle data. Represented as:

[0071]

[0072] The key generator frequency data under the k-th operating scenario are analyzed sequentially. Standardization was performed to obtain standardized key generator frequency data. Represented as:

[0073]

[0074] Among them, K s denoted as the number of running scenarios corresponding to the s-th stability label.

[0075] Optionally, in one embodiment of this application, a layer-by-layer training method based on a deep autoencoder, using simulation data for training, yields encoders and decoders corresponding to different stability labels, including:

[0076] The preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data for the scenario corresponding to the s-th stability label are sequentially selected from the simulation data to constitute the input feature data for the scenario corresponding to the s-th stability label. X is the input feature of the j-th running scenario with the s-th stability label, where the j-th running scenario with the s-th stability label corresponds to the j_s-th running scenario among the K running scenarios. j_s Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data

[0077] Define the encoder structure, where the encoder input data is X. j_s The output is a dimension w s eigenvector h s The encoder's output dimension and hidden layer settings are determined based on actual data and human experience;

[0078] Define the structure of the decoder, where the input data of the decoder is the encoder output of dimension w. s eigenvector h s The output is the dimension and X. j_s Same feature vector The hidden layer settings of the decoder are determined based on the characteristics of the actual data and human experience;

[0079] Based on the input data of all operating scenarios using the s-th stability label and the defined encoder and decoder, the encoder B corresponding to the s-th category label is trained using a layer-by-layer training method of a deep autoencoder and the constructed loss function loss. s and decoder C s The expression for the loss function is:

[0080]

[0081] Where D is the set of running scenario IDs participating in the training, d is the number of running scenarios in set D, and r is the r-th running scenario in set D.

[0082]

[0083] Where ε is the set threshold for the deviation between the predicted value and the true value of the feature.

[0084] Optionally, in one embodiment of this application, based on simulation data, encoders, and decoders corresponding to different stability labels, the maximum deviation output by the encoder and decoder for each type of stability label is obtained, including:

[0085] Will Input to the corresponding encoder B s and decoder C s Then, the output vector is obtained.

[0086] The maximum deviation of all scene samples corresponding to the s-th stability label after passing through the encoder-decoder output is calculated using the following formula:

[0087]

[0088] in,

[0089]

[0090] Optionally, in one embodiment of this application, simulation data from various operating scenarios of the power system are used as input data, and the text of manually analyzed results is used as output data to fine-tune the large model, resulting in a finely tuned curve analysis model, including:

[0091] Choose a large language model with input word vector dimension o;

[0092] The preprocessed simulation data is arranged into three-dimensional data according to the variable number dimension, time dimension, and variable category dimension. The dimension of this three-dimensional data is (2×S)×T×3.

[0093] Construct an input processing layer, which is used to reduce the dimensionality of the three-dimensional data to an o-dimensional vector, and the parameter to be learned in the input processing layer is W;

[0094] The preprocessed simulation data is used as input feature data and passed sequentially through the input processing layer and the large language model to obtain the output results. Based on the output results and the text of the manual analysis results, a loss function is constructed to fine-tune the learning parameters W of the input processing layer and the parameters of the large language model to obtain the fine-tuned curve analysis model.

[0095] Optionally, in one embodiment of this application, determining the analysis result of the power system simulation curve data to be analyzed by judging that the deviation between the input and output is less than the maximum deviation corresponding to its stability label includes:

[0096] Calculate the output vector of the decoder With input feature data (x1,...,x e ,...,x 6×S×T The deviation g of ) is expressed as:

[0097]

[0098] in,

[0099]

[0100] Compare the deviation g with the maximum deviation g corresponding to the z-th stability label. z The size of g, if g ≤ g z If the accuracy of the analysis result text is determined, the analysis result text will be used as the analysis result of the power system simulation curve data to be analyzed.

[0101] Optionally, in one embodiment of this application, the curve analysis module is further configured to:

[0102] If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, the analysis result text is judged to be inaccurate. The analysis result is obtained by manually analyzing the power system simulation curve data to be analyzed.

[0103] The power system simulation curve analysis method and system based on a large language model in this application introduces a large language model to convert the input simulation curve data into analysis result text, providing decision-making basis in text form; a method based on an autoencoder is designed to prompt for potentially erroneous results, ensuring that the simulation curve analysis results have a certain degree of accuracy and improving the practicality of simulation curve analysis applications.

[0104] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0105] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0106] Figure 1 This is a flowchart illustrating a power system simulation curve analysis method based on a large language model, as provided in Embodiment 1 of this application.

[0107] Figure 2 This is a schematic diagram of the structure of a power system simulation curve analysis system based on a large language model, provided in an embodiment of this application. Detailed Implementation

[0108] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote 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 intended to explain this application, and should not be construed as limiting this application.

[0109] The following describes, with reference to the accompanying drawings, a method and system for analyzing power system simulation curves based on a large language model, according to embodiments of this application.

[0110] Figure 1 This is a flowchart illustrating a power system simulation curve analysis method based on a large language model, as provided in Embodiment 1 of this application.

[0111] like Figure 1 As shown, the power system simulation curve analysis method based on a large language model includes the following steps:

[0112] Step 101: Obtain simulation data under various operating scenarios of the power system, and analyze the simulation data manually to obtain the manual analysis result text, which includes the stability label of the simulation data and the reason for the judgment.

[0113] Step 102: Based on the layer-by-layer training method of deep autoencoders, training is performed using simulation data to obtain encoders and decoders corresponding to different stability labels;

[0114] Step 103: Based on the simulation data, encoder and decoder corresponding to different stability labels, obtain the maximum deviation of the sample output by the encoder and decoder for each type of stability label;

[0115] Step 104: Using simulation data from various operating scenarios of the power system as input data and the text of manual analysis results as output data, fine-tune the large model to obtain the fine-tuned curve analysis model.

[0116] Step 105: Input the power system simulation curve data to be analyzed into the fine-tuned curve analysis model and output the analysis result text.

[0117] Step 106: Determine the stability label of the simulation curve data based on the analysis result text, and input the simulation curve data into the encoder and decoder corresponding to its stability label to obtain the output vector. Then, determine that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, and determine the analysis result of the power system simulation curve data to be analyzed.

[0118] Optionally, in one embodiment of this application, obtaining simulation data under various operating scenarios of the power system includes:

[0119] For a power system with N generators and M nodes, K operating scenarios are randomly set based on the historical operation and load forecast of the power system. The node voltage amplitude curve, generator rotor angle curve and generator frequency curve under the K operating scenarios are obtained by power flow calculation and transient simulation. In this embodiment, the selected power system has 10 generators and 39 nodes, and 20,000 operating conditions are randomly set.

[0120] The voltage amplitude data of the top S nodes with the largest and smallest average voltage amplitudes are selected sequentially from the node voltage amplitude curves to form the key node voltage amplitude data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T. If the number of generators in the curve is less than S, then the existing generators are selected repeatedly.

[0121] The rotor angle data of the top S generators with the largest and smallest average rotor angles are selected sequentially from the generator rotor angle curves to form the key generator rotor angle data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T. If the number of generators in the curve is less than S, then the existing generators are selected repeatedly.

[0122] The frequency data of the top S generators with the highest and lowest average frequencies are selected sequentially from the generator frequency curves to form the key generator frequency data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T. If the number of generators in the curve is less than S, then the existing generators are selected repeatedly.

[0123] In this embodiment, T is set to 1000 and S is set to 3.

[0124] Optionally, in one embodiment of this application, the simulation data is analyzed manually to obtain the manually analyzed result data, including:

[0125] Based on the safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability label and judgment reason as the manual analysis result text. Among them, there are a types of stability labels in the manual analysis result text. In this embodiment, the stability judgment result includes four categories: stable, transient power angle instability, transient voltage instability, and transient frequency instability, i.e., a=4.

[0126] Optionally, in one embodiment of this application, the simulation data needs to be preprocessed before training. The preprocessing process includes:

[0127] Sequentially analyze the key generator rotor angle data under the k-th operating scenario. Standardization was performed to obtain standardized key generator rotor angle data. Represented as:

[0128]

[0129] The key generator frequency data under the k-th operating scenario are analyzed sequentially. Standardization was performed to obtain standardized key generator frequency data. Represented as:

[0130]

[0131] Among them, K s denoted as the number of running scenarios corresponding to the s-th stability label.

[0132] Optionally, in one embodiment of this application, a layer-by-layer training method based on a deep autoencoder, using simulation data for training, yields encoders and decoders corresponding to different stability labels, including:

[0133] The preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data for the scenario corresponding to the s-th stability label are sequentially selected from the simulation data to constitute the input feature data for the scenario corresponding to the s-th stability label. X is the input feature of the j-th running scenario with the s-th stability label, where the j-th running scenario with the s-th stability label corresponds to the j_s-th running scenario among the K running scenarios. j_s Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data

[0134] Define the encoder structure, where the encoder input data is X. j_s The output is a dimension w s eigenvector h s The encoder's output dimension and hidden layer settings are determined based on actual data and human experience;

[0135] Define the structure of the decoder, where the input data of the decoder is the encoder output of dimension w. s eigenvector h s The output is the dimension and X. j_s Same feature vector The hidden layer settings of the decoder are determined based on the characteristics of the actual data and human experience;

[0136] Based on the input data of all operating scenarios using the s-th stability label and the defined encoder and decoder, the encoder B corresponding to the s-th category label is trained using a layer-by-layer training method of a deep autoencoder and the constructed loss function loss. s and decoder C s The expression for the loss function is:

[0137]

[0138] Where D is the set of running scenario IDs participating in the training, d is the number of running scenarios in set D, and r is the r-th running scenario in set D.

[0139]

[0140] The loss function described above is equivalent to setting a tolerance error range ε ​​for each feature, allowing the deviation between the predicted value and the true value of the feature to not exceed the set threshold ε. In this embodiment, ε is set to 0.01.

[0141] Optionally, in one embodiment of this application, based on simulation data, encoders, and decoders corresponding to different stability labels, the maximum deviation output by the encoder and decoder for each type of stability label is obtained, including:

[0142] Will Input to the corresponding encoder B s and decoder C s Then, the output vector is obtained.

[0143] The maximum deviation of all scene samples corresponding to the s-th stability label after passing through the encoder-decoder output is calculated using the following formula:

[0144]

[0145] in,

[0146]

[0147] Optionally, in one embodiment of this application, simulation data from various operating scenarios of the power system are used as input data, and manually analyzed data are used as output data to fine-tune the large model, resulting in a fine-tuned curve analysis model, including:

[0148] Choose a large language model whose input word vector dimension is O;

[0149] The preprocessed simulation data is arranged into three-dimensional data according to the variable number dimension, time dimension, and variable category dimension. The dimension of this three-dimensional data is (2×S)×T×3.

[0150] An input processing layer is constructed, which is used to reduce the dimensionality of the three-dimensional data to an o-dimensional vector. The parameter to be learned in the input processing layer is W. In this embodiment, a convolutional layer is selected to process the three-dimensional data and flatten it before inputting it into a fully connected layer.

[0151] The preprocessed simulation data is used as input feature data and passed sequentially through the input processing layer and the large language model to obtain the output result. Based on the output result and the manually analyzed text, a loss function is constructed to fine-tune the learning parameters W of the input processing layer and the parameters of the large language model to obtain the fine-tuned curve analysis model. The input of the large language model is the output vector of the input features after the input processing layer, and the output of the large language model is the manually analyzed text.

[0152] Optionally, in one embodiment of this application, determining the analysis result of the power system simulation curve data to be analyzed by judging that the deviation between the input and output is less than the maximum deviation corresponding to its stability label includes:

[0153] The stability label in the analysis results text is determined to be the z-th label, and the preprocessed simulation curve data is then... Let it be denoted as (x1,...,x e ,...,x 6×S×T ), and set (x1,...,x e ,...,x 6×S×T The input is fed into the encoder and decoder corresponding to the z-th stability label to obtain the output vector.

[0154] Calculate the output vector of the decoder With input feature data (x1,...,x e ,...,x 6×S×T The deviation g of ) is expressed as:

[0155]

[0156] in,

[0157]

[0158] Compare the deviation g with the maximum deviation g corresponding to the z-th stability label. z The size of g, if g ≤ g z If the judgment result of the large language model is correct, then the output result of the large language model is likely to be correct.

[0159] Optionally, in one embodiment of this application, the method further includes:

[0160] If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, i.e., g > g z The judgment results of the large language model may be inaccurate, suggesting that manual intervention is needed to analyze the curve.

[0161] The power system simulation curve analysis method and system based on a large language model in this application introduces a large language model to convert the input simulation curve data into analysis result text, providing decision-making basis in text form; a method based on an autoencoder is designed to prompt for potentially erroneous results, ensuring that the simulation curve analysis results have a certain degree of accuracy and improving the practicality of simulation curve analysis applications.

[0162] To achieve the above embodiments, this application also proposes a power system simulation curve analysis system based on a large language model.

[0163] Figure 2 This is a schematic diagram of the structure of a power system simulation curve analysis system based on a large language model, provided in an embodiment of this application.

[0164] like Figure 2 As shown, the power system simulation curve analysis system of this large language model includes:

[0165] The data acquisition module is used to acquire simulation data under various operating scenarios of the power system, and to obtain the manual analysis result text through manual analysis of the simulation data. The manual analysis result text includes the stability label of the simulation data and the reason for the judgment.

[0166] The encoder-decoder training module is used for training based on a layer-by-layer training method of deep autoencoders, using simulation data to obtain encoders and decoders corresponding to different stability labels.

[0167] The output deviation statistics module is used to obtain the maximum deviation of the sample corresponding to each stability label through the encoder and decoder based on the simulation data, encoder and decoder corresponding to different stability labels.

[0168] The curve analysis model training module is used to fine-tune the large model by taking simulation data from various operating scenarios of the power system as input data and text of manual analysis results as output data, and to obtain a fine-tuned curve analysis model.

[0169] The curve analysis module is used to input the simulated curve data of the power system to be analyzed into the finely tuned curve analysis model and output the analysis results text.

[0170] The curve analysis module is also used to determine the stability label of the simulation curve data based on the analysis result text, and input the simulation curve data into the encoder and decoder corresponding to its stability label to obtain the output vector. It also determines that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, and determines the analysis result of the power system simulation curve data to be analyzed.

[0171] Optionally, in one embodiment of this application, obtaining simulation data under various operating scenarios of the power system includes:

[0172] Based on the historical operation of the power system and load forecast, K operating scenarios are randomly set, and the node voltage amplitude curves, generator rotor angle curves and generator frequency curves under the K operating scenarios are obtained by power flow calculation and transient simulation calculation.

[0173] The voltage amplitude data of the top S nodes with the largest and smallest average voltage amplitudes are selected sequentially from the node voltage amplitude curves to form the key node voltage amplitude data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0174] The rotor angle data of the top S generators with the largest and smallest average rotor angles are selected sequentially from the generator rotor angle curves to form the key generator rotor angle data. Where n = 1, 2, ..., 2S, t is the t-th sampling time in the simulation curve, t = 1, ..., T;

[0175] The frequency data of the top S generators with the highest and lowest average frequencies are selected sequentially from the generator frequency curves to form the key generator frequency data. Where n = 1, 2, ..., 2S, and t is the t-th sampling time in the simulation curve, t = 1, ..., T.

[0176] Optionally, in one embodiment of this application, the simulation data is analyzed manually to obtain a text of the manual analysis results, including:

[0177] Based on the safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability labels and judgment reasons as the manual analysis result text. Among them, there are a types of stability labels in the manual analysis result text.

[0178] Optionally, in one embodiment of this application, the data acquisition module is further configured to preprocess the simulation data before training with the simulation data, the preprocessing process including:

[0179] Sequentially analyze the key generator rotor angle data under the k-th operating scenario. Standardization was performed to obtain standardized key generator rotor angle data. Represented as:

[0180]

[0181] The key generator frequency data under the k-th operating scenario are analyzed sequentially. Standardization was performed to obtain standardized key generator frequency data. Represented as:

[0182]

[0183] Among them, K s denoted as the number of running scenarios corresponding to the s-th stability label.

[0184] Optionally, in one embodiment of this application, a layer-by-layer training method based on a deep autoencoder, using simulation data for training, yields encoders and decoders corresponding to different stability labels, including:

[0185] The preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data for the scenario corresponding to the s-th stability label are sequentially selected from the simulation data to constitute the input feature data for the scenario corresponding to the s-th stability label. X is the input feature of the j-th running scenario with the s-th stability label, where the j-th running scenario with the s-th stability label corresponds to the j_s-th running scenario among the K running scenarios. j_s Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data

[0186] Define the encoder structure, where the encoder input data is X. j_s The output is a dimension w s eigenvector h s The encoder's output dimension and hidden layer settings are determined based on actual data and human experience;

[0187] Define the structure of the decoder, where the input data of the decoder is the encoder output of dimension w. s eigenvector h s The output is the dimension and X. j_s Same feature vector The hidden layer settings of the decoder are determined based on the characteristics of the actual data and human experience;

[0188] Based on the input data of all operating scenarios using the s-th stability label and the defined encoder and decoder, the encoder B corresponding to the s-th category label is trained using a layer-by-layer training method of a deep autoencoder and the constructed loss function loss. s and decoder C s The expression for the loss function is:

[0189]

[0190] Where D is the set of running scenario IDs participating in the training, d is the number of running scenarios in set D, and r is the r-th running scenario in set D.

[0191]

[0192] Where ε is the set threshold for the deviation between the predicted value and the true value of the feature.

[0193] Optionally, in one embodiment of this application, based on simulation data, encoders, and decoders corresponding to different stability labels, the maximum deviation output by the encoder and decoder for each type of stability label is obtained, including:

[0194] Will Input to the corresponding encoder B s and decoder C s Then, the output vector is obtained.

[0195] The maximum deviation of all scene samples corresponding to the s-th stability label after passing through the encoder-decoder output is calculated using the following formula:

[0196]

[0197] in,

[0198]

[0199] Optionally, in one embodiment of this application, simulation data from various operating scenarios of the power system are used as input data, and the text of manually analyzed results is used as output data to fine-tune the large model, resulting in a finely tuned curve analysis model, including:

[0200] Choose a large language model with input word vector dimension o;

[0201] The preprocessed simulation data is arranged into three-dimensional data according to the variable number dimension, time dimension, and variable category dimension. The dimension of this three-dimensional data is (2×S)×T×3.

[0202] Construct an input processing layer, which is used to reduce the dimensionality of the three-dimensional data to an o-dimensional vector, and the parameter to be learned in the input processing layer is W;

[0203] The preprocessed simulation data is used as input feature data and passed sequentially through the input processing layer and the large language model to obtain the output results. Based on the output results and the text of the manual analysis results, a loss function is constructed to fine-tune the learning parameters W of the input processing layer and the parameters of the large language model to obtain the fine-tuned curve analysis model.

[0204] Optionally, in one embodiment of this application, determining the analysis result of the power system simulation curve data to be analyzed by judging that the deviation between the input and output is less than the maximum deviation corresponding to its stability label includes:

[0205] Calculate the output vector of the decoder With input feature data (x1,...,x e ,...,x 6×S×T The deviation g of ) is expressed as:

[0206]

[0207] in,

[0208]

[0209] Compare the deviation g with the maximum deviation g corresponding to the z-th stability label. z The size of g, if g ≤ g z If the accuracy of the analysis result text is determined, the analysis result text will be used as the analysis result of the power system simulation curve data to be analyzed.

[0210] Optionally, in one embodiment of this application, the curve analysis module is further configured to:

[0211] If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, the analysis result text is judged to be inaccurate. The analysis result is obtained by manually analyzing the power system simulation curve data to be analyzed.

[0212] It should be noted that the explanation of the above-mentioned embodiment of the power system simulation curve analysis method of large language model also applies to the power system simulation curve analysis system of large language model in this embodiment, and will not be repeated here.

[0213] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0214] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0215] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0216] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0217] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0218] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0219] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0220] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for analyzing power system simulation curves based on a large language model, characterized in that, include: Simulation data under various operating scenarios of the power system is acquired, and the simulation data is analyzed manually to obtain a text of the manual analysis results. The text of the manual analysis results includes the stability label of the simulation data and the reason for the judgment. A layer-by-layer training method based on deep autoencoders is used to train on simulation data to obtain encoders and decoders corresponding to different stability labels. Based on the simulation data, encoder and decoder corresponding to different stability labels, the maximum deviation of the sample output by the encoder and decoder for each type of stability label is obtained. Simulation data from various operating scenarios of the power system are used as input data, and the text of manual analysis results is used as output data. The large model is then fine-tuned to obtain a fine-tuned curve analysis model. Input the simulated curve data of the power system to be analyzed into the fine-tuned curve analysis model, and output the analysis result text; Based on the analysis results text, the stability label of the simulation curve data is determined, and the simulation curve data is input into the encoder and decoder corresponding to its stability label to obtain the output vector. It is determined that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, and the analysis result of the power system simulation curve data to be analyzed is determined. The acquisition of simulation data under various operating scenarios of the power system includes: Based on the historical operating conditions of the power system and load forecasting, random settings are configured. In various operating scenarios, power flow calculations and transient simulations are used to obtain... The node voltage amplitude curve, generator rotor angle curve, and generator frequency curve under various operating scenarios; Select the nodes with the largest and smallest average voltage amplitudes sequentially from the node voltage amplitude curves. The voltage amplitude data of the nodes constitutes the voltage amplitude data of the key nodes. ,in, , For the first in the simulation curve Each sampling time, ; Select the generator rotor angles with the largest and smallest average rotor angles sequentially from the generator rotor angle curves. The rotor angle data of the generator constitutes the key generator rotor angle data. ,in, , For the first in the simulation curve Each sampling time, ; Select the generators with the highest and lowest average frequencies sequentially from the generator frequency curves. Generator frequency data constitutes key generator frequency data. ,in, , For the first in the simulation curve Each sampling time, ; The process of manually analyzing simulation data to obtain the resulting text includes: Based on safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability label and the judgment reasons as the manual analysis result text. This manual analysis result text contains a total of... Stability tags for different categories.

2. The method as described in claim 1, characterized in that, Before using the simulation data for training, the method further includes: preprocessing the simulation data, the preprocessing process including: sequentially for the first Key generator rotor angle data under various operating scenarios Standardization was performed to obtain standardized key generator rotor angle data. , is represented as: sequentially for the first Key generator frequency data under various operating scenarios Standardization was performed to obtain standardized key generator frequency data. , is represented as: in, For the first The number of running scenarios corresponding to each stability label.

3. The method as described in claim 2, characterized in that, The layer-by-layer training method based on deep autoencoders, using simulation data for training, obtains encoders and decoders corresponding to different stability labels, including: Select the first from the simulation data in sequence The stability label corresponds to the preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data in the corresponding scenario, constituting the first stability label. The stability labels correspond to the input feature data of the scene, where, For the first The first type of stability label The input features of the first running scenario, the first The first type of stability label Each running scenario corresponds to The first in the type of operation scenario One operating scenario, Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data ; Define the encoder structure, wherein the input data of the encoder is: The output is a dimension of eigenvectors The output dimension and hidden layer settings of the encoder are set based on actual data and human experience; Define the structure of the decoder, wherein the input data of the decoder is the encoder output in dimension . eigenvectors The output is the dimension and Same feature vector The hidden layer of the decoder is set according to the actual data characteristics and human experience; Based on the The input data for all operating scenarios of this stability label, along with the defined encoder and decoder, are used to train the deep autoencoder layer by layer using a constructed loss function. Training yielded the first Encoder corresponding to category label and decoder , where the loss function The expression is: Among them, set The set of serial numbers for the running scenarios participating in the training. For set The number of running scenarios in For set The first in One operating scenario, in, The threshold value for the deviation between the predicted and actual values ​​of a given feature.

4. The method as described in claim 3, characterized in that, The method, based on simulation data corresponding to different stability labels, encoders, and decoders, obtains the maximum deviation of the samples corresponding to each stability label through the encoder and decoder outputs, including: Will Input to the corresponding encoder and decoder Then, the output vector is obtained. ; Calculate all the first The maximum deviation of the scene samples corresponding to the stability label after passing through the encoder-decoder output is calculated using the following formula: in, 。 5. The method as described in claim 2, characterized in that, The process involves using simulation data from various power system operating scenarios as input data and textual analysis results as output data to fine-tune a large model, resulting in a finely tuned curve analysis model. This includes: Select the input word vector dimension as Large language models; The preprocessed simulation data is arranged into a three-dimensional dataset according to the variable number dimension, time dimension, and variable category dimension. The dimensions of this three-dimensional dataset are: ; An input processing layer is constructed, wherein the input processing layer is used to reduce the dimensionality of the three-dimensional data to... A dimensional vector, wherein the parameters to be learned in the input processing layer are... ; The preprocessed simulation data is used as input feature data, which is then passed through the input processing layer and the large language model to obtain the output results. Based on the output results and the text analysis results, a loss function is constructed to adjust the learning parameters of the input processing layer. The parameters of the large language model are fine-tuned to obtain a finely tuned curve analysis model.

6. The method as described in claim 4, characterized in that, The determination of the analysis results for the power system simulation curve data to be analyzed, based on the condition that the deviation between the input and output is less than the maximum deviation corresponding to its stability label, includes: Calculate the output vector of the decoder With input feature data deviation , is represented as: in, Comparison deviation With the Maximum deviation corresponding to the stability label The size, if If the accuracy of the analysis result text is determined, the analysis result text will be used as the analysis result of the power system simulation curve data to be analyzed.

7. The method as described in claim 1, characterized in that, The method further includes: If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, the analysis result text is judged to be inaccurate. The analysis result is obtained by manually analyzing the power system simulation curve data to be analyzed.

8. A power system simulation curve analysis system based on a large language model, characterized in that, include: The data acquisition module is used to acquire simulation data under various operating scenarios of the power system, and to obtain the manual analysis result text by manually analyzing the simulation data. The manual analysis result text includes the stability label of the simulation data and the reason for the judgment. The encoder-decoder training module is used for training based on a layer-by-layer training method of deep autoencoders, using simulation data to obtain encoders and decoders corresponding to different stability labels. The output deviation statistics module is used to obtain the maximum deviation of the sample corresponding to each stability label through the encoder and decoder based on the simulation data, encoder and decoder corresponding to different stability labels. The curve analysis model training module is used to fine-tune the large model by taking simulation data from various operating scenarios of the power system as input data and text of manual analysis results as output data, and to obtain a fine-tuned curve analysis model. The curve analysis module is used to input the simulated curve data of the power system to be analyzed into the finely tuned curve analysis model and output the analysis results text. The curve analysis module is also used to determine the stability label of the simulation curve data based on the analysis result text, input the simulation curve data into the encoder and decoder corresponding to its stability label to obtain the output vector, and determine that the deviation between the input data and the output vector is less than the maximum deviation corresponding to its stability label, thereby determining the analysis result of the power system simulation curve data to be analyzed. The acquisition of simulation data under various operating scenarios of the power system includes: Based on the historical operating conditions of the power system and load forecasting, random settings are configured. In various operating scenarios, power flow calculations and transient simulations are used to obtain... The node voltage amplitude curve, generator rotor angle curve, and generator frequency curve under various operating scenarios; Select the nodes with the largest and smallest average voltage amplitudes sequentially from the node voltage amplitude curves. The voltage amplitude data of the nodes constitutes the voltage amplitude data of the key nodes. ,in, , For the first in the simulation curve Each sampling time, ; Select the generator rotor angles with the largest and smallest average rotor angles sequentially from the generator rotor angle curves. The rotor angle data of the generator constitutes the key generator rotor angle data. ,in, , For the first in the simulation curve Each sampling time, ; Select the generators with the highest and lowest average frequencies sequentially from the generator frequency curves. Generator frequency data constitutes key generator frequency data. ,in, , For the first in the simulation curve Each sampling time, ; The process of manually analyzing simulation data to obtain the resulting text includes: Based on safety and stability guidelines, the node voltage amplitude curve, generator rotor angle curve, and generator frequency curve are analyzed to obtain the system stability label and the judgment reasons as the manual analysis result text. This manual analysis result text contains a total of... Stability tags for different categories.

9. The system as described in claim 8, characterized in that, The data acquisition module is further configured to preprocess the simulation data before training with the simulation data, the preprocessing process including: sequentially for the first Key generator rotor angle data under various operating scenarios Standardization was performed to obtain standardized key generator rotor angle data. , is represented as: sequentially for the first Key generator frequency data under various operating scenarios Standardization was performed to obtain standardized key generator frequency data. , is represented as: in, For the first The number of running scenarios corresponding to each stability label.

10. The system as described in claim 9, characterized in that, The layer-by-layer training method based on deep autoencoders, using simulation data for training, obtains encoders and decoders corresponding to different stability labels, including: Select the first from the simulation data in sequence The stability label corresponds to the preprocessed key generator rotor angle data, key node voltage amplitude data, and preprocessed key generator frequency data in the corresponding scenario, constituting the first stability label. The stability labels correspond to the input feature data of the scene, where, For the first The first type of stability label The input features of the first running scenario, the first The first type of stability label Each running scenario corresponds to The first in the type of operation scenario One operating scenario, Including preprocessed key generator rotor angle data Key bus voltage amplitude data Preprocessed key generator frequency data ; Define the encoder structure, wherein the input data of the encoder is: The output is a dimension of eigenvectors The output dimension and hidden layer settings of the encoder are set based on actual data and human experience; Define the structure of the decoder, wherein the input data of the decoder is the encoder output in dimension . eigenvectors The output is the dimension and Same feature vector The hidden layer of the decoder is set according to the actual data characteristics and human experience; Based on the The input data for all operating scenarios of this stability label, along with the defined encoder and decoder, are used to train the deep autoencoder layer by layer using a constructed loss function. Training yielded the first Encoder corresponding to category label and decoder , where the loss function The expression is: Among them, set The set of serial numbers for the running scenarios participating in the training. For set The number of running scenarios in For set The first in One operating scenario, in, The threshold value for the deviation between the predicted and actual values ​​of a given feature.

11. The system as claimed in claim 10, characterized in that, The method, based on simulation data corresponding to different stability labels, encoders, and decoders, obtains the maximum deviation of the samples corresponding to each stability label through the encoder and decoder outputs, including: Will Input to the corresponding encoder and decoder Then, the output vector is obtained. ; Calculate all the first The maximum deviation of the scene samples corresponding to the stability label after passing through the encoder-decoder output is calculated using the following formula: in, 。 12. The system as described in claim 9, characterized in that, The process involves using simulation data from various power system operating scenarios as input data and textual analysis results as output data to fine-tune a large model, resulting in a finely tuned curve analysis model. This includes: Select the input word vector dimension as Large language models; The preprocessed simulation data is arranged into a three-dimensional dataset according to the variable number dimension, time dimension, and variable category dimension. The dimensions of this three-dimensional dataset are: ; An input processing layer is constructed, wherein the input processing layer is used to reduce the dimensionality of the three-dimensional data to... A dimensional vector, wherein the parameters to be learned in the input processing layer are... ; The preprocessed simulation data is used as input feature data, which is then passed through the input processing layer and the large language model to obtain the output results. Based on the output results and the text analysis results, a loss function is constructed to adjust the learning parameters of the input processing layer. The parameters of the large language model are fine-tuned to obtain a finely tuned curve analysis model.

13. The system as described in claim 11, characterized in that, The determination of the analysis results for the power system simulation curve data to be analyzed, based on the condition that the deviation between the input and output is less than the maximum deviation corresponding to its stability label, includes: Calculate the output vector of the decoder With input feature data deviation , is represented as: in, Comparison deviation With the Maximum deviation corresponding to the stability label The size, if If the accuracy of the analysis result text is determined, the analysis result text will be used as the analysis result of the power system simulation curve data to be analyzed.

14. The system as described in claim 8, characterized in that, The curve analysis module is also used for: If the deviation between the input data and the output vector is greater than the maximum deviation corresponding to its stability label, the analysis result text is judged to be inaccurate. The analysis result is obtained by manually analyzing the power system simulation curve data to be analyzed.

Images