A method and system for ensuring safety incidents in power grid operation

Abstract

The present invention relates to the field of power grid monitoring technology, and more specifically to a method and system for ensuring safety incidents in power grid operation. The method comprises the following steps: obtaining a long-range dependency factor based on the autocorrelation attenuation characteristics of a monitoring data sequence; obtaining a periodic factor based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence; performing time series decomposition on the monitoring data sequence to obtain a residual term; obtaining a trend factor based on the fluctuation difference characteristics between the residual term and the monitoring data sequence; obtaining a nonlinear factor based on the nonlinear characteristics of the monitoring data sequence; obtaining fitness based on the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor; and obtaining a target monitoring data sequence based on the fitness of all monitoring data sequences. The method predicts the target monitoring data sequence using an LSTM model, analyzes the fault risk in the power grid based on the prediction results, and improves the accuracy of early warning of security incidents and the formulation of recovery plans after the incidents occur.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid monitoring, and in particular to a method and system for ensuring safety incidents in power grid operation. Background Art

[0002] In power grid operations, timely and effective handling of safety incidents is key to ensuring stable operation and reliable power supply. For example, safety incidents involving sudden and unusual events such as lightning strikes on power lines are generally handled through three key measures. The first is event detection and initial response. Line protection devices quickly detect abnormal changes in electrical quantities such as current and voltage. Meanwhile, fault recorders record the waveforms of electrical quantities before and after the fault occurs, providing detailed data for subsequent fault analysis. The second is fault isolation and current discharge protection. To prevent further expansion of the fault, the system immediately activates the parallel discharge gap device, which is connected in parallel to the line insulator. When an overvoltage occurs on the line, the discharge gap breaks down and discharges the lightning current and other overvoltage energy to the ground, reducing the risk of equipment damage and buying time for subsequent fault handling. The third is the formulation of emergency command and recovery plans for the control center. By collecting historical data provided by the SCADA data acquisition and monitoring control system and the EMS energy management system, the real-time operating status of the power grid is analyzed and predicted, and the impact of abnormal situations on the overall operation of the power grid, such as changes in power flow distribution and voltage stability, is evaluated. This measure can be combined with machine learning prediction models to analyze the real-time operating status of the power grid and evaluate the impact of faults based on the model prediction results. It plays an important role in providing early warning of security incidents and formulating recovery plans after the incidents.

[0003] SCADA and EMS are two key systems in circuit system automation. They collect real-time data from various sensors, instruments, and devices, such as temperature, voltage, current, and power. This type of analytical data is time-series data with periodic fluctuations and trends. Traditionally, this type of data is considered suitable for processing using the common LSTM long-short-term memory network model, which can effectively analyze time series data and capture the temporal changes in power flows. However, due to the large number of time data indicators to be analyzed, not all types of indicators are robust to LSTM models. This may be due to the data's inherent characteristics and the degree of interference with historical data. This results in low prediction accuracy for some monitoring data, ultimately affecting the accuracy of early warnings for security incidents and the development of recovery plans after incidents. Summary of the Invention

[0004] In order to solve the technical problem that the prediction accuracy of some of the monitoring data through the LSTM model is low, which affects the accuracy of early warning of security incidents and the formulation of recovery plans after the incident occurs, the purpose of the present invention is to provide a security incident protection method and system for power grid operation. The technical solutions adopted are as follows:

[0005] Obtain different types of monitoring data sequences from each monitoring node in the power grid;

[0006] Obtaining a long-range dependency factor based on the autocorrelation attenuation characteristics of the monitoring data sequence; obtaining a periodic factor based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence; performing time series decomposition on the monitoring data sequence to obtain a residual term; obtaining a trend factor based on the fluctuation difference characteristics between the residual term and the monitoring data sequence; and obtaining a nonlinear factor based on the nonlinear characteristics of the monitoring data sequence.

[0007] Obtaining fitness according to the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor; obtaining a target monitoring data sequence according to the fitness of all monitoring data sequences;

[0008] Predictions are made using the LSTM model based on the target monitoring data sequence, and the fault risks in the power grid are analyzed based on the prediction results.

[0009] Furthermore, the step of obtaining the long-range dependence factor according to the autocorrelation attenuation characteristics of the monitoring data sequence includes:

[0010] The seasonal term of the monitoring data sequence is removed by a time series decomposition algorithm to obtain a preprocessed sequence; the autocorrelation coefficient of the preprocessed sequence at a preset initial lag order is calculated to obtain a first value; when the autocorrelation coefficient corresponding to any lag order infinitely approaches half of the first value, the arbitrary lag order is normalized to obtain the long-range dependence factor.

[0011] Furthermore, the step of obtaining a periodic factor according to energy intensity distribution characteristics in the spectrum data of the monitoring data sequence includes:

[0012] The monitoring data sequence is subjected to discrete Fourier transform to obtain the spectrum data; the ratio of the maximum energy intensity in the spectrum data to the average energy intensity of all frequencies is calculated and normalized to obtain the periodic factor.

[0013] Furthermore, the step of obtaining a trend factor according to the fluctuation difference characteristics between the residual term and the monitoring data sequence includes:

[0014] The ratio of the standard deviation of the monitoring data sequence to the standard deviation of the residual term is calculated and normalized to obtain the trend factor.

[0015] Furthermore, the step of obtaining a nonlinear factor according to the nonlinear characteristics of the monitoring data sequence includes:

[0016] The Hurst exponent of the monitoring data sequence is calculated and negatively correlated to obtain the nonlinear factor.

[0017] Furthermore, the step of obtaining fitness according to the long-range dependence factor, the periodic factor, the trend factor and the nonlinear factor includes:

[0018] The average values ​​of the long-range dependence factor, the periodic factor, the trend factor and the nonlinear factor are calculated to obtain the fitness of the monitoring data sequence.

[0019] Furthermore, the step of obtaining fitness according to the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor; and obtaining a target monitoring data sequence according to the fitness of all monitoring data sequences includes:

[0020] All monitoring data sequences are sorted from large to small according to the fitness, and the monitoring data sequence that exceeds a preset position in the sorting is used as the target monitoring data sequence.

[0021] The present invention also proposes a security event protection system for power grid operation, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement any one of the steps of a security event protection method for power grid operation.

[0022] The present invention has the following beneficial effects:

[0023] In the present invention, a long-range dependency factor is obtained based on the autocorrelation attenuation characteristics of the monitoring data sequence, which can reflect whether the monitoring data sequence has a long-range dependency feature. A periodic factor is obtained based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence, which can reflect whether the monitoring data sequence has a periodic feature. A trend factor is obtained based on the fluctuation difference characteristics between the residual term and the monitoring data sequence, which can reflect whether the monitoring data sequence has a trend feature. A nonlinear factor is obtained based on the nonlinear characteristics of the monitoring data sequence, which can reflect whether the monitoring data sequence has a nonlinear change feature. Obtaining the fitness of the monitoring data sequence can comprehensively reflect whether it is suitable for prediction by the LSTM algorithm, thereby improving the accuracy of judgment. Obtaining the target monitoring data sequence from all monitoring data sequences can determine the range suitable for prediction by the LSTM algorithm, and eliminating the monitoring data sequences that are not suitable for the algorithm, so that the credibility of the model prediction results is higher. The power grid control center can more accurately analyze and evaluate the power grid operation status and fault impact based on more accurate prediction results, thereby improving the accuracy of early warning of security incidents and the formulation of recovery plans after the incident. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A flow chart of a method for ensuring safety incidents in power grid operations provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0026] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method and system for ensuring safety incidents in power grid operations according to the present invention, including its specific implementation, structure, features, and effectiveness. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0027] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0028] The following describes in detail a method and system for ensuring safety incidents in power grid operation provided by the present invention with reference to the accompanying drawings.

[0029] See also Figure 1 , which shows a flow chart of a method for ensuring security events in power grid operation according to an embodiment of the present invention, the method comprising the following steps:

[0030] Step S1: Acquire different types of monitoring data sequences of each monitoring node in the power grid.

[0031] Different types of monitoring data sequences are obtained for each monitoring node in the power grid. In the embodiment of the present invention, the monitoring objects include active power, reactive power, current, voltage amplitude and phase angle, temperature of each line node, and generator operating status parameters at the power generation end. The implementer can determine the monitoring objects according to the implementation scenario. The collected data are cleaned and preprocessed. For missing values ​​in the data, interpolation is used to fill them. For example, if the current data of a monitoring node at a certain moment is missing, linear interpolation is performed based on the data before and after to obtain an approximate value. All collected time series data are normalized to the maximum and minimum values ​​to eliminate the dimensional effect and facilitate subsequent analysis. After the data processing is completed, different types of monitoring data sequences are obtained for each monitoring node. The collection range of the monitoring data sequence is historical data within a period of time before the safety accident occurs. The implementer can determine it according to the implementation scenario.

[0032] Step S2, obtaining a long-range dependence factor based on the autocorrelation attenuation characteristics of the monitoring data sequence; obtaining a periodic factor based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence; performing time series decomposition on the monitoring data sequence to obtain a residual term; obtaining a trend factor based on the fluctuation difference characteristics between the residual term and the monitoring data sequence; and obtaining a nonlinear factor based on the nonlinear characteristics of the monitoring data sequence.

[0033] As a specialized recurrent neural network, the LSTM (Long Short-Term Memory) network effectively processes time series data. It is particularly well-suited for capturing the changing patterns of power flow distribution across different lines in current power grid security incident analysis. Analysis of the LSTM algorithm reveals that it is particularly well-suited for time series data with long-range dependencies, strong periodicity, and trends. However, due to inherent data characteristics and the degree of historical data interference, some types of monitoring data series are unsuitable for predictive analysis using this algorithm, resulting in low prediction accuracy. This impacts both early warning of security incidents and the accuracy of post-incident recovery plans. Therefore, it is necessary to identify, among all acquired monitoring data series, those that offer the highest confidence in the output of predictions using the LSTM algorithm.

[0034] Furthermore, traditional RNN recurrent neural networks have problems with gradient explosion and gradient vanishing when processing long sequences, resulting in an inability to effectively capture long-range dependency features. LSTM, however, uses a gate control mechanism designed through its unique network design structure to effectively solve the gradient problem and is more suitable for time series data with long-range dependencies. Therefore, a long-range dependency factor can be obtained based on the autocorrelation decay characteristics of the monitoring data sequence. Preferably, in an embodiment of the present invention, the step of obtaining the long-range dependency factor includes: removing the seasonal term of the monitoring data sequence through a time series decomposition algorithm to obtain a preprocessed sequence. It should be noted that the time series decomposition algorithm belongs to the prior art, and the specific steps will not be repeated. The preprocessed sequence after removing the seasonal term can avoid the influence of the periodic characteristics of the monitoring data sequence on the acquisition of the long-range dependency factor, thereby improving the accuracy of the long-range dependency factor. The autocorrelation coefficient of the preprocessed sequence at a preset initial lag order is calculated to obtain a first value. It should be noted that the calculation of the autocorrelation coefficient belongs to the prior art, and the specific steps will not be repeated. The autocorrelation coefficient measures the degree of correlation of the same event at different times. When the autocorrelation coefficient corresponding to any lag order approaches infinitely to half of the first value, the arbitrary lag order is normalized to obtain the long-range dependence factor; wherein the normalization is to perform maximum and minimum normalization on the arbitrary lag order corresponding to all monitoring data sequences. It should be noted that the preset initial lag order is 1. When the arbitrary lag order is larger, it means that the half-life of the autocorrelation coefficient of the monitoring data sequence is longer. As the lag order increases, the autocorrelation coefficient slowly decreases. The more obvious the long-range dependence characteristics of the monitoring data sequence, the larger the long-range dependence factor, and the more suitable the monitoring data sequence is for prediction using LSTM. The formulas for obtaining the long-range dependence factor include:

[0035]

[0036] Where, Indicates the autocorrelation coefficient when the lag order is x. Indicates the preset initial lag order, represents the first value, U represents any lag order, represents the normalization of any lag order, and C represents the long-range dependence factor.

[0037] LSTM dynamically adjusts information weights through a gating mechanism, effectively adapting to the periodic patterns of time series data. At the same time, the cumulative nature of the memory unit enables it to capture long-term trends in the data. Therefore, the LSTM algorithm is more suitable for monitoring data time series with periodic and trend characteristics, and has higher prediction accuracy. Therefore, a periodic factor can be obtained based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence. Preferably, in an embodiment of the present invention, the step of obtaining the periodic factor includes: performing a discrete Fourier transform on the monitoring data sequence to obtain spectrum data. It should be noted that discrete Fourier transform belongs to the prior art, and the specific steps are not repeated here. Each frequency in the spectrum data corresponds to an energy intensity value, and the ratio of the maximum energy intensity in the spectrum data to the average energy intensity of all frequencies is calculated and normalized to obtain the periodic factor. This normalization is maximum / minimum normalization. A larger periodic factor means that the maximum energy intensity in the spectrum data is higher than the average level, the more pronounced the peak feature is, and the more significant the periodic characteristics of the monitoring data sequence are, making the monitoring data sequence more suitable for prediction using LSTM.

[0038] Furthermore, after obtaining the periodic factor of the monitoring data sequence, it is necessary to determine whether there is a trend feature, so the monitoring data sequence is decomposed into a time series to obtain a residual term; and a trend factor is obtained based on the fluctuation difference characteristics of the residual term and the monitoring data sequence. Preferably, in an embodiment of the present invention, the step of obtaining the trend factor includes: calculating the ratio of the standard deviation of the monitoring data sequence to the standard deviation of the residual term and normalizing it to obtain a trend factor; the normalization is maximum and minimum value normalization. The reason why the decomposed residual term is analyzed instead of the trend term in the process of calculating the trend factor is that: the trend part after decomposition is usually a low-frequency smooth curve, usually obtained based on moving average or local weighted regression, and its fluctuation amplitude may be artificially smoothed, making it difficult to directly reflect the true strength of the trend; and the residual feature can reflect the quality of the trend fitting. Since the residual term is the remaining part after subtracting the trend term and the seasonal term from the original monitoring data sequence, if the trend is the main component of the data, the residual fluctuation should be small. If the trend is weak or does not exist, the residual fluctuation is more obvious; secondly, the residual term is insensitive to outliers, thereby reducing the impact of outliers, while directly analyzing the trend term is easily distorted by extreme values, resulting in low accuracy. If the trend characteristics of the monitoring data sequence are more obvious, the residual fluctuation characteristics will be weaker. At this time, the standard deviation of the residual term will be smaller than the standard deviation of the original monitoring data sequence. The larger the trend factor, the more obvious the trend characteristics of the monitoring data sequence, and thus the monitoring data sequence will be more suitable for prediction using LSTM.

[0039] The input gate, forget gate, and output gate of LSTM all implement nonlinear transformations through sigmoid functions and tanh functions, which can capture the complex nonlinear dependency characteristics of time series data. LSTM can dynamically adjust nonlinear influences through a gating mechanism, making the algorithm more suitable for time series data with nonlinear characteristic relationships in the monitoring data sequence. Therefore, a nonlinear factor is obtained based on the nonlinear characteristics of the monitoring data sequence; preferably, in an embodiment of the present invention, the step of obtaining the nonlinear factor includes: calculating the Hurst exponent of the monitoring data sequence and negatively mapping it to obtain the nonlinear factor; it should be noted that the Hurst exponent belongs to the prior art and is usually calculated using the rescaled range analysis method. The specific steps will not be repeated here. When the Hurst exponent is smaller, it means that the anti-persistence of the monitoring data sequence is stronger, the nonlinear characteristics are stronger, and the more suitable it is for the LSTM algorithm to make predictions.

[0040] Step S3, obtaining fitness according to the long-range dependency factor, the period factor, the trend factor and the nonlinear factor; obtaining the target monitoring data sequence according to the fitness of all monitoring data sequences.

[0041] After obtaining the long-range dependency factor, periodic factor, trend factor and nonlinear factor of the monitoring sequence, the fitness can be obtained according to the long-range dependency factor, periodic factor, trend factor and nonlinear factor; preferably, in an embodiment of the present invention, the step of obtaining the fitness includes: calculating the average value of the long-range dependency factor, periodic factor, trend factor and nonlinear factor to obtain the fitness of the monitoring data sequence. The greater the fitness, the more suitable the monitoring data sequence is for prediction by the LSTM algorithm, and the higher the accuracy of the prediction result; the smaller the fitness, the lower the accuracy of the monitoring data sequence for prediction using the LSTM algorithm. Further, the monitoring data sequence used for LSTM prediction can be determined by the fitness, so the target monitoring data sequence is obtained according to the fitness of all monitoring data sequences; preferably, in an embodiment of the present invention, the step of obtaining the target monitoring data sequence includes: sorting all monitoring data sequences from large to small according to the fitness, and taking the monitoring data sequence before the preset position in the sorting as the target monitoring data sequence. If the fitness of the overall monitoring data sequence is greater, the preset rank value can be adjusted accordingly to increase the number of target monitoring data sequences, and if the fitness of the overall monitoring data sequence is low, the preset rank value can be adjusted accordingly to decrease the number of target monitoring data sequences to ensure the credibility of the prediction model results and improve the prediction efficiency. In an embodiment of the present invention, the preset rank is the median of the sorting quantity. The target monitoring data sequence can be predicted by the LSTM model to obtain a more accurate prediction result.

[0042] Step S4: perform predictions using the LSTM model based on the target monitoring data sequence, and analyze the fault risks in the power grid based on the prediction results.

[0043] After obtaining the target monitoring data sequence from all monitoring nodes in the power grid, the target monitoring data sequence can be predicted using an LSTM, allowing the grid's operating status to be analyzed and the fault impact to be assessed based on the prediction results. First, a training set and a test set are constructed for the target monitoring data sequence. The training set is used to train the model's parameters, and the test set is used to evaluate the model's performance. During the training process, the LSTM model's hyperparameters are adjusted to optimize the model's performance, ensuring that the model's prediction accuracy meets the required performance. It should be noted that the LSTM model is state-of-the-art, and the specific training steps are not detailed here. After training is complete, the target monitoring data sequence is input over a period of time to obtain the predicted value of the target monitoring data at the next moment. Once the model for the target monitoring data sequence is constructed, security events can be predicted and analyzed. Based on the model's prediction results, the fault impact can be assessed. For example, if the model's prediction results indicate a significant increase in the active power of a critical line, exceeding its rated capacity, and the grid control center determines the impact to be severe based on the prediction results, this indicates that the line is at risk of overload, impacting the grid's power supply reliability. The control center then needs to develop a plan to ensure the smooth operation of the line and reduce the fault's impact on the grid. At this point, by analyzing whether each monitoring node is suitable for the LSTM model, the appropriate data type is selected for prediction, which improves the prediction accuracy of monitoring data in the power grid, avoids errors in prediction results when the model is not applicable, and ultimately improves the accuracy of early warning of security incidents and the formulation of recovery plans after incidents.

[0044] In summary, an embodiment of the present invention provides a method for ensuring safety incidents in power grid operation; a long-range dependency factor is obtained based on the autocorrelation attenuation characteristics of a monitoring data sequence; a periodic factor is obtained based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence; the monitoring data sequence is time-series decomposed to obtain a residual term; a trend factor is obtained based on the fluctuation difference characteristics between the residual term and the monitoring data sequence; a nonlinear factor is obtained based on the nonlinear characteristics of the monitoring data sequence; fitness is obtained based on the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor; and a target monitoring data sequence is obtained based on the fitness of all monitoring data sequences. The present invention predicts the target monitoring data sequence through an LSTM model, analyzes the fault risk in the power grid based on the prediction results, and improves the accuracy of early warning of safety incidents and the formulation of recovery plans after the incident occurs.

[0045] The present invention also proposes a security incident protection system for power grid operation, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement any one of the steps of a security incident protection method for power grid operation.

[0046] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0047] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for ensuring safety incidents in power grid operation, characterized in that: The method comprises the following steps: Obtain different types of monitoring data sequences from each monitoring node in the power grid; Obtaining a long-range dependency factor based on the autocorrelation attenuation characteristics of the monitoring data sequence; obtaining a periodic factor based on the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence; performing time series decomposition on the monitoring data sequence to obtain a residual term; obtaining a trend factor based on the fluctuation difference characteristics between the residual term and the monitoring data sequence; and obtaining a nonlinear factor based on the nonlinear characteristics of the monitoring data sequence. Obtaining fitness according to the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor; obtaining a target monitoring data sequence according to the fitness of all monitoring data sequences; Predicting the target monitoring data sequence using an LSTM model, and analyzing the fault risk in the power grid based on the prediction results; the step of obtaining the long-range dependency factor based on the autocorrelation attenuation characteristics of the monitoring data sequence includes: The seasonal term of the monitoring data sequence is removed by a time series decomposition algorithm to obtain a preprocessed sequence; the autocorrelation coefficient of the preprocessed sequence at a preset initial lag order is calculated to obtain a first value; when the autocorrelation coefficient corresponding to any lag order infinitely approaches half of the first value, the arbitrary lag order is normalized to obtain the long-range dependence factor; the step of obtaining the periodic factor according to the energy intensity distribution characteristics in the spectrum data of the monitoring data sequence includes: The monitoring data sequence is subjected to a discrete Fourier transform to obtain the spectrum data; the ratio of the maximum energy intensity in the spectrum data to the average energy intensity of all frequencies is calculated and normalized to obtain the period factor; the step of obtaining the trend factor based on the fluctuation difference characteristics between the residual term and the monitoring data sequence includes: The ratio of the standard deviation of the monitoring data sequence to the standard deviation of the residual term is calculated and normalized to obtain the trend factor; the step of obtaining the nonlinear factor according to the nonlinear characteristics of the monitoring data sequence includes: Calculating the Hurst index of the monitoring data sequence and performing negative correlation mapping to obtain the nonlinear factor; and obtaining fitness according to the long-range dependency factor, the periodic factor, the trend factor, and the nonlinear factor comprises: The average values ​​of the long-range dependence factor, the periodic factor, the trend factor and the nonlinear factor are calculated to obtain the fitness of the monitoring data sequence.

2. A method for ensuring safety incidents in power grid operation according to claim 1, characterized in that: Obtaining fitness according to the long-range dependency factor, the period factor, the trend factor and the nonlinear factor; The steps for obtaining the target monitoring data sequence according to the fitness of all monitoring data sequences include: All monitoring data sequences are sorted from large to small according to the fitness, and the monitoring data sequence that exceeds a preset position in the sorting is used as the target monitoring data sequence.

3. A security incident protection system for power grid operation, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 2.

Images