Security situation assessment method, system, equipment and medium

Abstract

The invention relates to the technical field of power distribution network monitoring, and discloses a security situation assessment method, system and device and a medium, and the method comprises the steps: obtaining load data, and dividing the load data into regular nodes and irregular nodes; calculating the load prediction value of the regular node, constructing and training a neural network model according to a calculation result, and predicting the load prediction value of the irregular node; according to the load prediction values of the regular nodes and the irregular nodes, calculating to obtain a power grid load prediction value, and optimizing the power grid load prediction value to obtain a first power grid load prediction value; and constructing and training a convolutional neural network model by using the power grid load predicted value, and inputting the first power grid load predicted value into the trained convolutional neural network model to obtain an evaluation result of the state of the power distribution network, thereby improving the operation safety and efficiency of the power grid, and enhancing the adaptability of safety situation analysis in a dynamic change environment.

Description

Technical Field

[0001] This invention relates to the field of power distribution network monitoring technology, and in particular to a method, system, equipment and medium for security status assessment. Background Technology

[0002] With the rapid development of smart grid technology, the monitoring and safety assessment of the operation status of distribution networks has become a crucial link in modern power systems. Traditional distribution network monitoring mainly relies on manual inspections and single electrical measuring devices. These methods are not only inefficient but also have certain monitoring blind spots. With the integration of sensor technology, Internet of Things (IoT), data mining, and artificial intelligence (AI) technologies, intelligent distribution network monitoring systems are gradually being applied. These systems integrate multiple sensor devices (such as voltage, current, and power sensors) and big data analysis methods to acquire real-time operation data of the distribution network and use advanced algorithms to process and analyze the data to perform tasks such as load forecasting, fault detection, and status assessment, thereby improving the operation efficiency and safety of the distribution network.

[0003] Existing power distribution network monitoring technologies still have some shortcomings. When dealing with load changes under complex power system conditions, they often fail to fully consider the real-time status changes of the power distribution network, thus failing to reflect the fluctuations in power demand in real time, which in turn affects the assessment of the power grid's security status. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a security situation assessment method, system, device, and medium to address the problem that when dealing with load changes in complex power system states, the real-time situation changes of the distribution network are often not fully considered, thus failing to reflect fluctuations in power demand in real time, which in turn affects the security situation assessment of the power grid.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a security situation assessment method, comprising: Obtain load data and divide the load data into regular nodes and irregular nodes; The load forecast values ​​for the regular nodes are calculated, and a neural network model is constructed and trained based on the calculation results to predict the load forecast values ​​for the irregular nodes. Based on the load forecast values ​​of the regular nodes and the irregular nodes, the power grid load forecast value is calculated, and the power grid load forecast value is optimized to obtain the first power grid load forecast value. A convolutional neural network model is constructed and trained using the power grid load forecast values. The first power grid load forecast value is then input into the trained convolutional neural network model to obtain the evaluation results of the distribution network status.

[0007] As a preferred embodiment of the security situation assessment method described in this invention, dividing the load data into regular nodes and irregular nodes includes: choose Each load data point is used as the initial cluster center; Calculate load data to The distance between cluster centers is used to assign the load data to the nearest cluster center; Recalculated after each allocation Identify cluster centers, and stop iterating when the change in cluster centers is less than a first threshold. Clustering results; In the clustering results, the cluster center is defined as the parent node, otherwise it is a child node; Calculate the deviation between the child node and the parent node; If the deviation is less than the second threshold, it is a regular node; otherwise, it is an irregular node.

[0008] The beneficial effects of this preferred technical solution are that, through clustering and deviation analysis, regular and irregular nodes can be accurately divided, thereby improving the accuracy of load forecasting and the safety of power grid operation.

[0009] As a preferred embodiment of the security situation assessment method described in this invention, the calculation of the load prediction value of the regular node includes: The ratio of regular child nodes to parent nodes is defined as the initial load allocation factor, and the ratio of the initial load allocation factor to the load data of regular child nodes is defined as the regular child node load allocation factor. Calculate the actual load values ​​of the child nodes and the parent nodes according to the calculation results, and then calculate the time weight. The dynamic load allocation factor is calculated using the regular sub-node load allocation factor and time weight.

[0010] As a preferred embodiment of the security situation assessment method described in this invention, wherein: All actual load values ​​are decomposed to obtain training trend data; The regression coefficients of the linear fit are calculated based on the training trend data, the trend line is fitted using an optimization algorithm, and the predicted load value of the parent node is calculated. The predicted load value of the regular child node is calculated based on the dynamic load allocation factor and the predicted load value of the parent node. This is the predicted load value of the regular node.

[0011] The beneficial effects of this preferred technical solution are that, through dynamic load allocation factors and trend fitting, it can accurately predict the load of regular nodes, thereby improving prediction accuracy and power grid operation efficiency.

[0012] As a preferred embodiment of the security situation assessment method described in this invention, the predicted load forecast values ​​for irregular nodes include: Obtain historical load data for irregular nodes and divide the historical load data into time series data; The time series data is mapped using Fourier transform to obtain the spectral energy distribution; Based on the spectral energy distribution, the spectral sensitivity is calculated using the spectral gradient change method. Spectral sensitivities greater than or equal to the third threshold are selected to obtain a set of sensitive frequencies. The sensitivity frequency set is calculated, and the calculation results are used to construct a load feature vector; Historical workload data is obtained, and the long short-term memory network model is trained using the historical workload data to obtain the trained long short-term memory network model. By inputting the load feature vector into the trained long short-term memory network model, the predicted load values ​​of irregular nodes are obtained.

[0013] The beneficial effects of this preferred technical solution are that, through spectrum analysis and long short-term memory network models, it can accurately predict irregular node loads, thereby enhancing the adaptability of the power grid and the accuracy of prediction.

[0014] As a preferred embodiment of the security situation assessment method described in this invention, optimizing the power grid load forecast includes: Based on the predicted load values ​​of regular nodes and irregular nodes, calculate the allocation weights for the predicted load values ​​of regular nodes and irregular nodes respectively. The load forecast values ​​of regular and irregular nodes are weighted and summed to obtain the power grid load forecast value. Obtain the rated voltage value of the power distribution network; The voltage deviation is obtained by calculating the difference between the voltage data in the load data and the rated voltage value. The situation factor is calculated based on the voltage deviation and the rated voltage value; By using situation factors, the power grid load forecast is corrected to obtain the first power grid load forecast.

[0015] In a preferred embodiment of the security situation assessment method described in this invention, the assessment results of the distribution network status include: The assessment results of the distribution network status are judged. If the assessment results of the distribution network status are greater than or equal to the fourth threshold, it is an abnormal state, and maintenance personnel are notified. If the assessment result of the distribution network status is less than the fourth threshold, it is considered to be in a normal state, and the assessment result of the distribution network status will continue to be monitored.

[0016] Secondly, the present invention provides a security situation assessment system, comprising: The partitioning module is used to acquire load data and divide the load data into regular nodes and irregular nodes; The prediction module is used to calculate the load prediction value of the regular nodes, construct and train a neural network model based on the calculation results, and predict the load prediction value of the irregular nodes. The calculation module is used to calculate the power grid load forecast value based on the load forecast values ​​of the regular nodes and the irregular nodes, and to optimize the power grid load forecast value to obtain the first power grid load forecast value. The evaluation module is used to construct and train a convolutional neural network model using the power grid load forecast value, and input the first power grid load forecast value into the trained convolutional neural network model to obtain the evaluation result of the distribution network status.

[0017] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the security situation assessment method described above.

[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the security situation assessment method described above.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention collects and preprocesses load data, dividing the preprocessed load data into regular and irregular nodes; calculates the load prediction value of regular nodes, constructs a Long Short-Term Memory (LSTM) network model to predict the load prediction value of irregular nodes; calculates and optimizes the power grid load prediction value based on the load prediction values ​​of regular and irregular nodes to obtain the final power grid load prediction value; and constructs a Convolutional Neural Network (CNN) model to conduct security assessment and monitoring of the final power grid load prediction value; thereby improving the operational safety and efficiency of the power grid and enhancing the adaptability of security situation assessment in dynamically changing environments. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the overall process logic of a security situation assessment method provided in one embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figure 1 As an embodiment of the present invention, a security situation assessment method is provided, comprising: S100: Acquire load data and divide the load data into regular nodes and irregular nodes; S200: Calculate the load forecast values ​​for regular nodes, construct and train a neural network model based on the calculation results, and predict the load forecast values ​​for irregular nodes. S300: Based on the load forecast values ​​of regular and irregular nodes, calculate the power grid load forecast value, optimize the power grid load forecast value, and obtain the first power grid load forecast value; S400: Convolutional neural network model is constructed and trained using the power grid load forecast value. The first power grid load forecast value is input into the trained convolutional neural network model to obtain the evaluation result of the distribution network status.

[0024] It should be noted that, firstly, regular and irregular nodes are identified to accurately pinpoint load characteristics; then, differentiated forecasting methods are employed for different nodes to improve forecast accuracy; next, high-precision power grid load forecasts are obtained through optimization and integration; finally, convolutional neural networks (CNNs) are used for state assessment, real-time anomaly monitoring, significantly enhancing the safety and efficiency of power grid operation and adapting to dynamically changing environments. In this embodiment of the invention, step S100 includes the following sub-steps A1-A6; In A1: Select Each load data point is used as the initial cluster center; In A2: Calculate load data to The distance between cluster centers is used to assign the load data to the nearest cluster center; In A3: Recalculate after each allocation. Identify cluster centers, and stop iterating when the change in cluster centers is less than a first threshold. Clustering results; In A4: the cluster centers in the clustering results are defined as parent nodes, otherwise as child nodes; In A5: Calculate the deviation between the child node and the parent node; In A6: if the deviation is less than the second threshold, it is a regular node; otherwise, it is an irregular node.

[0025] Specifically, the voltage, current and power data of the power distribution network are collected in real time by using smart sensors. The smart sensors include voltage, current and power sensors, and the load data includes voltage, current and power data. Preprocessing of the load data includes using the standard deviation method to identify and remove outlier data, using Lagrange interpolation to fill in missing data, using wavelet to denoise the load data, and then standardizing the denoised load data.

[0026] It should be noted that by using intelligent sensors to collect voltage, current, and power data of the distribution network in real time, accurate monitoring of the distribution network's operating status is achieved. Abnormal data is identified and removed using the standard deviation method, effectively improving data quality and ensuring the accuracy of subsequent load forecasting models. The standard deviation method is simple, computationally inexpensive, and can operate efficiently in real-time data processing, making it suitable for real-time monitoring of large-scale distribution networks. Lagrange interpolation is used to fill in missing data, ensuring the integrity and continuity of load data. Wavelet denoising technology, through multi-scale signal analysis, effectively removes noise components while retaining useful information in the load data. Standardized data eliminates dimensional differences, allowing data with different characteristics to be effectively compared within the same model. This not only improves the comparability of load data but also enhances the adaptability of the forecasting model under different environments, increasing the data's universality and scalability.

[0027] In this embodiment of the invention, the preprocessed load data is divided into regular nodes and irregular nodes; Specifically, an empirical method is used to set the first threshold, for example, the first threshold could be 0.01; the elbow rule is used to set the number of clusters. ; Randomly select from preprocessed load data Initial cluster centers for the load data were determined, and the Euclidean distance was used to calculate the preprocessed load data. The Euclidean distance between each cluster center is used to assign the preprocessed load data to the nearest cluster center. The calculation is repeated after each assignment. The algorithm calculates cluster centers, and stops iterating when the number of calculated cluster centers is less than a first threshold, thus obtaining the classified results. Each clustering result generates a node set; In the clustering results, a cluster is defined as a parent node, representing the load pattern of that cluster; otherwise, it is defined as a child node. The dynamic threshold, or second threshold, is defined using statistical analysis methods and is expressed as follows: , in, The second threshold, This represents the mean Mahalanobis distance between child nodes and their parent nodes. The parameter, representing the standard deviation from the mean, determines the stringency of the threshold. is the standard deviation of the Mahalanobis distance between all child nodes and their parent nodes; The deviation between child nodes and parent nodes is calculated using Mahalanobis distance. The deviation is then compared with a second threshold. If the deviation is less than or equal to the second threshold, the node is considered a regular node. If the deviation is greater than the second threshold, the node is considered an irregular node.

[0028] It should be noted that cluster analysis determines the number of clusters using the elbow rule and assigns data points to the nearest cluster centers using the Euclidean distance method. This effectively divides load data into different categories. Through automated clustering, manual intervention can be reduced, and intelligent analysis of large-scale distribution network load data can be achieved. Using statistical analysis methods to set dynamic thresholds, the classification criteria can be automatically adjusted according to the actual distribution of the data. By calculating the deviation between child nodes and parent nodes using Mahalanobis distance, the covariance between data points can be taken into account, accurately measuring the differences and improving the accuracy of load data classification. This allows load data to be more finely divided into regular and irregular nodes, greatly enhancing the adaptability of the system.

[0029] In this embodiment of the invention, step S200 includes the following sub-steps B1-A6; In B1: the ratio of regular child nodes to parent nodes is defined as the initial load allocation factor, and the ratio of the initial load allocation factor to the load data of regular child nodes is defined as the regular child node load allocation factor. In B2: Calculate the actual load value of the child node and the actual load value of the parent node, and calculate the time weight based on the calculation results; In B3: The dynamic load allocation factor is calculated using the regular sub-node load allocation factor and time weight.

[0030] Specifically, training data is collected from the UCI machine learning library based on the pattern nodes and preprocessed. The ratio of pattern child nodes to parent nodes is defined as the initial load allocation factor. , The index for the node set links the regular child node load data with the initial load allocation factor. The ratio is defined as the regular child node load distribution factor. , For node set The index of the regular child node in the text; Based on the preprocessed load data and the current and voltage data in the training data, the active power is calculated using the power formula and defined as the actual load value of the regular sub-node. The mean of the actual load values ​​of the regular sub-node is calculated and defined as the actual load value of the parent node. The time weight is calculated using the exponential decay weighting method. Based on the regular sub-node load allocation factor and time weight, the dynamic load allocation factor is calculated as follows: , in, As time weight, For regular child node load allocation factors, For time windows.

[0031] In this embodiment of the invention, after completing steps B1-B3, step S200 also includes steps B4-B6. In B4: All actual load values ​​are decomposed to obtain training trend data; In B5: The regression coefficients of the linear fit are calculated based on the training trend data, the trend line is fitted using an optimization algorithm, and the predicted load value of the parent node is calculated. In B6: The predicted load value of the regular child node is calculated based on the dynamic load allocation factor and the predicted load value of the parent node, which is the predicted load value of the regular node.

[0032] Specifically, the moving average method is used to decompose all actual load values ​​to obtain training trend data, and the regression coefficients for linear fitting are calculated based on the training trend data. and , is represented as: , , in, For time indexing, To train trend data, To train the mean of trend data, The average of the time indices. For training trend data; Based on regression coefficients and The trend line was fitted using the least squares method, and the calculation was performed over time. The predicted load value of the parent node is expressed as: , in, For time; Based on the dynamic load allocation factor and the predicted load value of the parent node, the predicted load value of the regular child node is calculated as follows: , in, For dynamic load allocation factor, The predicted load value for the parent node; The predicted load value of the regular child node Defined as the load forecast value of regular nodes; It should be noted that by collecting and preprocessing current and voltage data from the UCI machine learning library, the quality of the data upon which load forecasting relies is ensured. By defining child node load allocation factors and time weights, the load relationship between parent and child nodes is explicitly quantified, enabling the load forecasting model to accurately reflect the load demand of each node. Utilizing the calculation of initial and dynamic load allocation factors, fine-tuning of load forecasting can be achieved, especially when dealing with complex power grid structures, ensuring that the load forecast for each node better matches actual needs. In calculating the dynamic changes in load, an exponential decay weighting method is used to adjust historical load data. By weighting the data, more recent data takes on a larger proportion in load forecasting, enhancing the model's responsiveness to short-term load fluctuations and improving the timeliness of forecasts. Combined with training trend data, the calculation of linear fitting regression coefficients further enhances the accuracy of the load forecasting model, providing a reliable basis for power grid dispatching and load management. The dynamic adjustment capability makes the power grid operation more efficient, enabling timely optimization of power grid dispatching based on changes in load demand, reducing energy waste, and improving the stability and reliability of the power grid. Calculating regression coefficients through linear fitting methods allows for the establishment of mathematical models for load forecasting based on historical data, which has significant academic value and practical significance.

[0033] In this embodiment of the invention, after completing steps B4-B6, step S200 above also includes steps B7-B12; In B7: Obtain historical load data for irregular nodes and divide the historical load data into time series data; In B8: The time series data is mapped using Fourier transform to obtain the spectral energy distribution; In B9: Based on the spectral energy distribution, the spectral sensitivity is calculated using the spectral gradient change method. Spectral sensitivities greater than or equal to the third threshold are selected to obtain a set of sensitive frequencies. In B10: the sensitivity frequency set is calculated, and the calculation results are used to construct a load feature vector; In B11: acquire historical load data, use the historical load data to train the long short-term memory network model, and obtain the trained long short-term memory network model. In B12: The load feature vector is input into the trained long short-term memory network model to obtain the load prediction value of irregular nodes.

[0034] Specifically, the fractional Fourier transform is used to map the time series data from the time domain to the fractional domain to obtain the spectral energy distribution; The sensitivity threshold, i.e. the third threshold, is set using the statistical percentile method. Based on the spectral energy distribution, the spectral sensitivity at the FrFT order is calculated using the spectral gradient change method. Spectral sensitivities greater than or equal to the third threshold are selected to generate a set of sensitive frequencies. The average frequency and mean square frequency of the sensitive frequency set are calculated using statistical analysis. The total energy of the sensitive frequencies is then calculated based on the sensitive frequency set and expressed as follows: , in, For the order of the fractional Fourier transform Spectral energy at the lower sensitive frequency For frequency index, For the set of sensitive frequencies, Total energy at the sensitive frequency; Construct a load characteristic vector from the average frequency, mean square frequency, and total energy at the sensitive frequency. Represented as: ,in, The average frequency, The mean square frequency, Total energy at the sensitive frequency; The historical load feature vector is calculated from the collected historical and real-time load data. and load eigenvector ; Construct a Long Short-Term Memory (LSTM) network model, including an input layer, hidden layers, and an output layer; The input layer is set as the load feature vector, and the long short-term memory network model is trained using the historical load feature vector. The model parameters are iteratively optimized using the loss function and the Adam optimizer. By inputting the load feature vector into the trained long short-term memory network model, the predicted load values ​​of irregular nodes are obtained. It should be noted that dividing load data into time series data using the sliding window method can extract local features of the data, helping the model identify short-term load fluctuations and periodic changes. For irregular nodes, the sliding window method can effectively capture irregular load changes in the short term, enabling the LSTM model to learn the time dependence in the load data more accurately. Using fractional Fourier transform to map time series data from the time domain to the fractional domain can effectively improve the ability to extract frequency features. The FrFT order can find a more suitable balance between the time domain and the frequency domain, thereby identifying the potential patterns of load data at different orders. By setting a sensitivity threshold and screening out important spectra, the model's response speed and prediction accuracy to key load changes can be effectively improved. The total energy of the sensitive frequency, as a feature quantity, is directly related to the load volatility and can help the prediction model better capture the dynamic trend of load changes. By combining features in the time domain and frequency domain, the load prediction model's ability to learn multi-level data patterns is improved. The LSTM model can more accurately identify the patterns of load changes from these features and predict the future load situation of irregular nodes.

[0035] In this embodiment of the invention, step S300 includes the following sub-steps C1-C6; In C1: Calculate the allocation weights of the load forecast values ​​for regular and irregular nodes based on the load forecast values ​​for regular and irregular nodes, respectively; In C2: The load forecast values ​​of regular and irregular nodes are weighted and summed to obtain the power grid load forecast value; In C3: Obtain the rated voltage value of the distribution network; In C4: Calculate the difference between the voltage data in the load data and the rated voltage value to obtain the voltage deviation; In C5: The situation factor is calculated based on the voltage deviation and the rated voltage value; In C6: The power grid load forecast is corrected using the situation factor to obtain the first power grid load forecast.

[0036] Specifically, based on the load forecast values ​​of regular nodes and irregular nodes, the entropy weight method is used to calculate the allocation weights of the load forecast values ​​of regular nodes and irregular nodes respectively. The preliminary power grid load forecast value is obtained by using a weighted summation based on the allocation weights of all the load forecast values ​​of regular nodes and irregular nodes. The rated voltage values ​​of the distribution network are collected according to IEC standards, and the voltage deviation is calculated based on the difference between the voltage data in the load data and the rated voltage value. The situation factor is calculated based on the voltage deviation and the rated voltage value, and is expressed as follows: , in, For voltage deviation, This is the rated voltage value; Based on the situation factors, the preliminary power grid load forecast is dynamically corrected to obtain the final power grid load forecast, expressed as: , in, These are preliminary power grid load forecasts. As a situation factor, This represents the final predicted grid load. It should be noted that the introduction of the entropy weight method enables the system to better adapt to the complexity and dynamism of different power grids, exhibiting stronger adaptability and versatility. The weighted summation method ensures that the preliminary power grid load forecast fully reflects the load characteristics of different nodes, avoiding errors that may arise from a single forecast value, and improving the stability and accuracy of the forecast results. Especially when facing complex power grid systems with large load fluctuations, it can effectively improve the reliability of the forecast. Adhering to IEC standards helps improve the safety and efficiency of power grid operation and enhances competitiveness in the international electricity market. By monitoring voltage deviations in real time, the power grid load can be dynamically adjusted, thereby reducing the overload risk of the power system and ensuring the stable operation of the power grid. Through timely feedback on the power grid operating status, the situation factor helps the forecast model quickly adapt to changes in the power grid, improving the real-time performance and accuracy of the forecast results, enhancing the accuracy of power grid load forecasting, and strengthening the model's dynamic adjustment capability. This allows power grid managers to adjust power grid operation strategies in a timely manner based on the forecast results, preventing overload or insufficient power supply, and improving the safety and efficiency of power grid operation.

[0037] In this embodiment of the invention, step S400 includes the following sub-steps D1-D2; In D1: The evaluation result of the distribution network status is judged. If the evaluation result of the distribution network status is greater than or equal to the fourth threshold, it is an abnormal status, and maintenance personnel are notified. In D2: If the assessment result of the distribution network status is less than the fourth threshold, it is in a normal state, and the assessment result of the power grid status continues to be monitored.

[0038] Specifically, a convolutional neural network model is constructed, including an input layer, convolutional layers, pooling layers, and a fully connected output layer; The input layer is defined as the power grid load forecast value. The convolutional neural network (CNN) model is trained using the dataset, and the model parameters are iteratively optimized using the loss function and the Adam optimizer. The final power grid load forecast is input into the trained convolutional neural network (CNN) model to obtain the evaluation results of the distribution network status. The quartile method is used to set the inspection threshold, which is the fourth threshold. The quartile method sorts the dataset of evaluation results of the distribution network status, calculates the interquartile range, sets the normal range as the upper and lower bounds, and selects the upper bound as the fourth threshold. The assessment result of the distribution network status is compared with the fourth threshold. If the assessment result of the distribution network status is greater than or equal to the fourth threshold, it is judged as an abnormal state and maintenance personnel are notified by email. If the assessment result of the distribution network status is less than the fourth threshold, it is judged as a normal state and the assessment result of the distribution network status continues to be monitored.

[0039] It should be noted that using labeled datasets from the UCI machine learning library helps ensure the high quality and reliability of the model's training data. The input layer is defined as the predicted grid load value. Convolutional layers can automatically extract local features from the data, which is particularly effective for time-series grid load data. Pooling layers can reduce the dimensionality of the data, avoid overfitting, and enhance the model's generalization ability. The fully connected output layer is responsible for transforming the high-order features extracted by the convolutional and pooling layers into the final evaluation result. Through this multi-level, staged data processing, CNNs can uncover deep-seated patterns from complex grid load data, improving prediction accuracy. By adaptively adjusting the learning rate, the Adam optimizer can adjust itself flexibly when faced with tasks of varying complexity, avoiding instability caused by an excessively large learning rate or slow convergence caused by an excessively small learning rate. It links the prediction results with the actual operating status of the distribution network, ensuring that the prediction of the grid load can reflect the true situation in a timely manner. The advantage of the quartile method lies in its simplicity and sensitivity to data distribution. It can automatically adapt to changes in different datasets, reduce the subjectivity of manually setting thresholds, and ensure that relevant personnel can be warned in the shortest possible time when grid anomalies occur, thereby reducing losses and improving the reliability and safety of grid operation.

[0040] The above is an illustrative scheme of a security situation assessment method according to this embodiment. It should be noted that the technical solution of this security situation assessment system and the technical solution of the aforementioned security situation assessment method belong to the same concept. Details not described in detail in the technical solution of the security situation assessment system in this embodiment can be found in the description of the technical solution of the aforementioned security situation assessment method.

[0041] The security situation assessment system in this embodiment includes: The partitioning module is used to acquire load data and divide the load data into regular nodes and irregular nodes; The prediction module is used to calculate the load prediction value of the regular nodes, construct and train a neural network model based on the calculation results, and predict the load prediction value of the irregular nodes. The calculation module is used to calculate the power grid load forecast value based on the load forecast values ​​of the regular nodes and the irregular nodes, and to optimize the power grid load forecast value to obtain the first power grid load forecast value. The evaluation module is used to construct and train a convolutional neural network model using the power grid load forecast value, and input the first power grid load forecast value into the trained convolutional neural network model to obtain the evaluation result of the distribution network status.

[0042] This embodiment also provides a computer device suitable for security situation assessment, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a security posture assessment method as described in the above embodiments.

[0043] This embodiment also provides a storage medium on which a computer program is stored, which, when executed by a processor, implements a security situation assessment method as proposed in the above embodiments.

[0044] The storage medium proposed in this embodiment and the security situation assessment method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0045] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0046] Example 2, referring to Table 1, differs from the first example and provides a verification test of the security situation assessment method to verify and explain the technical effects used in this method.

[0047] By comparing the performance of traditional methods with that of this method in distribution network load forecasting and security assessment, the advantages of this method in forecasting accuracy, computation time and security assessment accuracy are verified. Traditional method: Selects ARIMA model, based solely on time series prediction, without dynamic correction and CNN evaluation; Prediction accuracy, computation time, and security assessment accuracy will be used as evaluation metrics. The simulated power distribution network data includes 500 nodes, 24-hour load data, and 10% abnormal conditions. The experimental simulation data is shown in Table 1. Table 1 Experimental Simulation Data Table , As shown in Table 1, the prediction accuracy of the traditional method is 0.72, indicating that the predicted value deviates significantly from the actual value, especially in the prediction of irregular nodes, because it cannot capture nonlinear changes. The traditional method takes 150 seconds to compute because it requires iterative calculation point by point and lacks parallel optimization. As for accuracy, the traditional method relies on the load prediction value and does not combine deep learning models for state assessment, resulting in a high false negative rate. The prediction accuracy of this method is 0.38, which is a significant improvement. Since the LSTM model is good at handling nonlinear time series with irregular nodes, dynamic correction further reduces the error through situation factors. Moreover, the time consumption of this method is less than that of traditional methods, thanks to the efficient parallel computing capabilities of LSTM and CNN, as well as the low complexity of dynamic correction. The accuracy of this method is much higher than that of traditional methods, thanks to the CNN model extracting deep features of the load data through convolution operations, and combined with quartile thresholds, it can more accurately identify abnormal states. Therefore, by combining the LSTM model to process nonlinear time series of irregular nodes, dynamically correcting and optimizing prediction accuracy, and using the CNN model to extract deep features for security assessment, the accuracy and efficiency of distribution network load forecasting are significantly improved, while the accuracy of abnormal state detection is enhanced.

[0048] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A security situation assessment method, characterized in that, include: Obtain load data and divide the load data into regular nodes and irregular nodes; The load forecast values ​​for the regular nodes are calculated, and a neural network model is constructed and trained based on the calculation results to predict the load forecast values ​​for the irregular nodes. Based on the load forecast values ​​of the regular nodes and the irregular nodes, the power grid load forecast value is calculated, and the power grid load forecast value is optimized to obtain the first power grid load forecast value. A convolutional neural network model is constructed and trained using the power grid load forecast values. The first power grid load forecast value is then input into the trained convolutional neural network model to obtain the evaluation results of the distribution network status.

2. The security situation assessment method as described in claim 1, characterized in that, Dividing load data into regular nodes and irregular nodes includes: choose Each load data point is used as the initial cluster center; Calculate load data to The distance between cluster centers is used to assign the load data to the nearest cluster center; Recalculated after each allocation Identify cluster centers, and stop iterating when the change in cluster centers is less than a first threshold. Clustering results; In the clustering results, the cluster center is defined as the parent node, otherwise it is a child node; Calculate the deviation between the child node and the parent node; If the deviation is less than the second threshold, it is a regular node; otherwise, it is an irregular node.

3. The security situation assessment method as described in claim 2, characterized in that, The calculation of the load forecast value for the regular nodes includes: The ratio of regular child nodes to parent nodes is defined as the initial load allocation factor, and the ratio of the initial load allocation factor to the load data of regular child nodes is defined as the regular child node load allocation factor. Calculate the actual load values ​​of the child nodes and the parent nodes according to the calculation results, and then calculate the time weight. The dynamic load allocation factor is calculated using the regular sub-node load allocation factor and time weight.

4. The security situation assessment method as described in claim 3, characterized in that, Also includes: All actual load values ​​are decomposed to obtain training trend data; The regression coefficients of the linear fit are calculated based on the training trend data, the trend line is fitted using an optimization algorithm, and the predicted load value of the parent node is calculated. The predicted load value of the regular child node is calculated based on the dynamic load allocation factor and the predicted load value of the parent node. This is the predicted load value of the regular node.

5. The security situation assessment method as described in claim 4, characterized in that, The predicted load forecasts for irregular nodes include: Obtain historical load data for irregular nodes and divide the historical load data into time series data; The time series data is mapped using Fourier transform to obtain the spectral energy distribution; Based on the spectral energy distribution, the spectral sensitivity is calculated using the spectral gradient change method. Spectral sensitivities greater than or equal to the third threshold are selected to obtain a set of sensitive frequencies. The sensitivity frequency set is calculated, and the calculation results are used to construct a load feature vector; Historical workload data is obtained, and the long short-term memory network model is trained using the historical workload data to obtain the trained long short-term memory network model. By inputting the load feature vector into the trained long short-term memory network model, the predicted load values ​​of irregular nodes are obtained.

6. The security situation assessment method as described in claim 5, characterized in that, Optimizing power grid load forecasts includes: Based on the predicted load values ​​of regular nodes and irregular nodes, calculate the allocation weights for the predicted load values ​​of regular nodes and irregular nodes respectively. The load forecast values ​​of regular and irregular nodes are weighted and summed to obtain the power grid load forecast value. Obtain the rated voltage value of the power distribution network; The voltage deviation is obtained by calculating the difference between the voltage data in the load data and the rated voltage value. The situation factor is calculated based on the voltage deviation and the rated voltage value; By using situation factors, the power grid load forecast is corrected to obtain the first power grid load forecast.

7. The security situation assessment method as described in claim 6, characterized in that, The assessment results of the distribution network status include: The assessment results of the distribution network status are judged. If the assessment results of the distribution network status are greater than or equal to the fourth threshold, it is an abnormal state, and maintenance personnel are notified. If the assessment result of the distribution network status is less than the fourth threshold, it is considered to be in a normal state, and the assessment result of the distribution network status will continue to be monitored.

8. A security situation assessment system, employing a security situation assessment method as described in any one of claims 1-7, characterized in that, include: The partitioning module is used to acquire load data and divide the load data into regular nodes and irregular nodes; The prediction module is used to calculate the load prediction value of the regular nodes, construct and train a neural network model based on the calculation results, and predict the load prediction value of the irregular nodes. The calculation module is used to calculate the power grid load forecast value based on the load forecast values ​​of the regular nodes and the irregular nodes, and to optimize the power grid load forecast value to obtain the first power grid load forecast value. The evaluation module is used to construct and train a convolutional neural network model using the power grid load forecast value, and input the first power grid load forecast value into the trained convolutional neural network model to obtain the evaluation result of the distribution network status.

9. A computer device, comprising: A memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of a security situation assessment method according to any one of claims 1 to 7.

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

Images