Power system abnormal load point identification method and device, computer equipment and medium product

Abstract

The invention relates to a power system abnormal load point identification method and device, computer equipment and a medium product. The method comprises the following steps: obtaining power load data of a power system, obtaining a load time sequence according to the power load data, carrying out detrending processing on the load time sequence to obtain a target time sequence, and carrying out unsupervised learning on the long-short-term memory network under the condition that the long-short-term memory network meets unsupervised learning conditions. The abnormal scores of the plurality of load data points are acquired through the long short-term memory network, the abnormal data points in the plurality of load data points are identified according to the abnormal scores, and the abnormal probability of each load data point is acquired through the optimized long short-term memory network under the condition that the long short-term memory network satisfies a supervised learning condition. And identifying abnormal data points in the plurality of load data points according to the abnormal probability. By adopting the method, the load abnormity can be accurately identified.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device, computer equipment, and media product for identifying abnormal load points in a power system. Background Art

[0002] As a key indicator of power system operation, monitoring and analyzing power load is fundamental to ensuring safe and reliable operation. Accurately understanding the dynamic changes in power load can help power dispatching departments rationally plan power generation, optimize grid operations, avoid power outages caused by overloads or abnormal fluctuations, and improve power supply quality.

[0003] Traditional methods mainly use methods based on empirical formulas and simple statistical analysis to monitor power loads. They usually rely on summarizing historical data, such as calculating indicators such as average load and maximum load, and judging whether the load is abnormal based on manual experience.

[0004] However, as the scale of power systems continues to expand, their structures become increasingly complex, and their load types become increasingly diverse, traditional methods have difficulty processing large-scale, high-dimensional load data, and are unable to effectively capture the complex patterns and subtle features in load changes, resulting in the inability to accurately identify abnormal loads. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, and medium product for identifying abnormal load points in a power system that can accurately identify load abnormalities in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for identifying abnormal load points in a power system, comprising:

[0007] Obtaining power load data of the power system and obtaining a load time series based on the power load data;

[0008] Detrending the load time series to obtain a target time series; the target time series includes load data corresponding to multiple load data points;

[0009] When the long short-term memory network meets the unsupervised learning conditions, anomaly scores of multiple load data points are obtained through the long short-term memory network, and abnormal data points among the multiple load data points are identified according to the anomaly scores;

[0010] When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and the abnormal data points among multiple load data points are identified according to the abnormal probability.

[0011] In one embodiment, the step of performing detrending processing on the load time series to obtain the target time series includes:

[0012] The load time series is fitted with a linear function to obtain a first detrended sequence, and the load time series is fitted with a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series;

[0013] Obtain the target time series according to the load time series, the first detrended series, and the second detrended series.

[0014] In one embodiment, the step of identifying abnormal data points among the plurality of load data points according to the abnormality scores includes:

[0015] Fitting normal distribution to the abnormal scores of multiple load data points;

[0016] For each load data point after normal distribution fitting, obtain the probability density value of the load data point under the normal distribution, and convert the probability density value into the outlier probability;

[0017] Identify outlier data points among multiple load data points based on outlier probabilities.

[0018] In one embodiment, the method further comprises:

[0019] For a time series group composed of multiple target time series, obtain the anomaly score corresponding to the load data point in each target time series in the time series group;

[0020] Under the target dimension, sum the anomaly scores corresponding to all target time series in the time series group to obtain the anomaly score corresponding to the time series group;

[0021] Identify abnormal data points in a time series group based on the anomaly scores corresponding to the time series group.

[0022] In one embodiment, the optimized long short-term memory network includes a long short-term memory layer, a fully connected layer and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0023] In one embodiment, a process of identifying an abnormal data point among a plurality of load data points includes:

[0024] When the outlier probability of the load data point exceeds a first threshold, or the abnormal probability of the load data point exceeds a second threshold, the load data point is determined to be an abnormal data point.

[0025] In a second aspect, the present application further provides a device for identifying abnormal load points in a power system, comprising:

[0026] A sequence acquisition module is used to acquire power load data of the power system and obtain a load time series based on the power load data;

[0027] A detrending processing module is used to perform detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to multiple load data points;

[0028] a first identification module, configured to obtain anomaly scores of multiple load data points through the long short-term memory network when the long short-term memory network satisfies unsupervised learning conditions, and identify abnormal data points among the multiple load data points based on the anomaly scores;

[0029] The second identification module is used to obtain the abnormal probability of each load data point through the optimized long short-term memory network when the long short-term memory network meets the supervised learning conditions, and identify abnormal data points among multiple load data points according to the abnormal probability.

[0030] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any one of the method steps in the first aspect when executing the computer program.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements any one of the method steps in the first aspect when the computer program is executed by a processor.

[0032] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which implements any one of the method steps in the first aspect when executed by a processor.

[0033] The above-mentioned method, device, computer equipment, and media product for identifying abnormal load points in a power system obtain the power load data of the power system, obtain the load time series based on the power load data, detrend the load time series, and obtain the target time series. When the long short-term memory network meets the unsupervised learning conditions, the abnormal scores of multiple load data points are obtained through the long short-term memory network, and abnormal data points among the multiple load data points are identified based on the abnormal scores. When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and abnormal data points among the multiple load data points are identified based on the abnormal probability. The method can be applied to abnormality identification in various scenarios, improve the comprehensiveness of abnormality identification, and thus accurately identify load abnormalities in the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 This is a diagram of an application environment of a method for identifying abnormal load points in a power system according to an embodiment;

[0036] Figure 2 1 is a flow chart of a method for identifying abnormal load points in a power system according to an embodiment;

[0037] Figure 3 is a structural block diagram of a power system abnormal load point identification system in one embodiment;

[0038] Figure 4 is a flow chart of a method for identifying abnormal load points in a power system according to another embodiment;

[0039] Figure 5 is a structural block diagram of a device for identifying abnormal load points in a power system in one embodiment;

[0040] Figure 6 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0042] The method for identifying abnormal load points in a power system provided by the embodiment of the present application can be applied to Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process. The data storage system can be integrated with server 104, or placed on a cloud or other network server. Terminal 102 is used to obtain power load data from the power system, obtain a load time series based on the power load data, detrend the load time series to obtain a target time series, and, if the long short-term memory network meets unsupervised learning conditions, obtain anomaly scores for multiple load data points using the long short-term memory network, and identify anomaly data points from the multiple load data points based on the anomaly scores. If the long short-term memory network meets supervised learning conditions, obtain anomaly probabilities for each load data point using the optimized long short-term memory network, and identify anomaly data points from the multiple load data points based on the anomaly probabilities. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart car devices, projectors, and the like. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0043] In an exemplary embodiment, Figure 2 As shown in the figure, a method for identifying abnormal load points in a power system is provided. Figure 1 Taking the terminal 102 in the example as an example, the following steps 202 to 208 are included. Among them:

[0044] S202: Obtain power load data of the power system, and obtain a load time series based on the power load data.

[0045] Optionally, the load time series is a time series data, which is formed by sampling and sorting the power load of the power system in chronological order to form sequence data with time dependence, so as to more effectively extract time characteristics.

[0046] S204: Detrending the load time series to obtain a target time series; the target time series includes load data corresponding to a plurality of load data points.

[0047] Optionally, detrending includes removing linear and cyclical trends. A linear trend refers to a monotonically increasing or decreasing trend in a time series, while a cyclical trend refers to regular fluctuations that recur at fixed intervals in a time series. These fluctuations are typically expressed mathematically as sine or cosine functions. Linear and cyclical trends are regular signals that can mask sudden changes in anomaly detection or short-term forecasting. When performing anomaly detection using time series models such as Long Short-Term Memory (LSTM), this can cause the LSTM to mistakenly identify short-term abnormal fluctuations as part of a trend, failing to identify the anomaly. Therefore, detrended data better conforms to the stationary assumptions of time series models such as LSTM, preventing the model from learning incorrect time dependencies due to trend terms.

[0048] S206: When the long short-term memory network satisfies the unsupervised learning condition, abnormality scores of the multiple load data points are obtained through the long short-term memory network, and abnormal data points among the multiple load data points are identified according to the abnormality scores.

[0049] Optionally, unsupervised learning conditions refer to training a long short-term memory network through unsupervised learning. Under unsupervised learning, the LSTM is trained using normal load data to learn normal patterns in the data. During anomaly identification, the target time series is input into the trained LSTM, and the reconstruction error (the difference between the input and output) is calculated. This reconstruction error is used as an anomaly score, with a higher score indicating a more likely anomaly.

[0050] S208: When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and abnormal data points among the multiple load data points are identified according to the abnormal probability.

[0051] Optionally, supervised learning conditions refer to training a long short-term memory network through supervised learning. Under supervised learning, the LSTM is trained using historical anomaly samples and corresponding labels to learn the anomaly types of the data. During anomaly identification, a fully connected layer is connected to the end of the LSTM, the output dimension is set, and the model's loss function is modified to a binary cross-entropy method. This results in the model outputting a sequence of 0-1 results, which is used to determine whether each data point is an anomaly.

[0052] In the above-mentioned method for identifying abnormal load points in a power system, the power load data of the power system is obtained, and the load time series is obtained based on the power load data, and the load time series is detrended to obtain a target time series. When the long short-term memory network meets the unsupervised learning conditions, the abnormal scores of multiple load data points are obtained through the long short-term memory network, and abnormal data points among the multiple load data points are identified based on the abnormal scores. When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and abnormal data points among the multiple load data points are identified based on the abnormal probability. This method can be applicable to abnormality identification in a variety of scenarios, improve the comprehensiveness of abnormality identification, and thus accurately identify load abnormalities in the power system.

[0053] In an exemplary embodiment, a load time series is detrended to obtain a target time series, including: fitting the load time series through a linear function to obtain a first detrended sequence, and fitting the load time series through a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; and the target time series is obtained according to the load time series, the first detrended sequence, and the second detrended sequence.

[0054] Optionally, to obtain a reliable long-range correlation series, the trend of the series needs to be separated from the original data (i.e., the load time series). Detrending a time series is a method for extracting temporal features of a time series. It is primarily used to detect long-term correlations in a time series and can extract long-range correlation information, thereby helping deep learning models more accurately capture temporal features. The main purpose of the detrending operation is to effectively remove the impact of strong trends in the original time series on the overall time series, so as to prevent strong trends from interfering with the model's learning effect. Typically, time features include linear and cyclical trends. The load time series is fitted with linear and sinusoidal functions, respectively, to remove the linear and cyclical trends, respectively.

[0055] For example, the mathematical expression of the linear function is:

[0056]

[0057] in, It represents the fitted value of the linear trend, that is, the part of the load time series that changes monotonically over a long period of time; x represents the time variable; k represents the slope of the linear trend, which reflects the growth / decay rate of the load over time; b represents the intercept of the linear trend.

[0058] The mathematical expression of the sine function is:

[0059]

[0060] in, It represents the fitting value of the periodic trend, that is, the part of the load time series that fluctuates periodically; x represents the time variable; represents the amplitude of the sine function; Related to the periodic frequency, it indicates the density of the sinusoidal curve; Represents the phase shift of the sine function; Indicates the vertical offset of the sine function.

[0061] Optionally, when removing the trend, the linear fitting value and the periodic fitting value are subtracted from the load time series. The resulting target time series is the series without the linear and periodic regularities, so that the model focuses more on abnormal fluctuations.

[0062] In this embodiment, the load time series is fitted by a linear function to obtain a first detrended sequence, and the load time series is fitted by a sine function to obtain a second detrended sequence. The target time series is obtained based on the load time series, the first detrended sequence and the second detrended sequence, which can ensure that the model accurately captures the time characteristics, thereby improving the accuracy of anomaly recognition.

[0063] In an exemplary embodiment, the step of identifying abnormal data points among multiple load data points based on abnormal scores includes: fitting the abnormal scores of the multiple load data points with a normal distribution; for each load data point after the normal distribution fitting, obtaining the probability density value of the load data point under the normal distribution, and converting the probability density value into an outlier probability; and identifying abnormal data points among the multiple load data points based on the outlier probability.

[0064] For example, in unsupervised learning, anomaly detection is performed based on the normal distribution outlier probability detection method. The anomaly detection method based on the normal distribution outlier probability is a common statistical method that assumes that the data conforms to the normal distribution and uses the properties of the normal distribution to estimate the degree of outliers of the data points. Assuming that the data conforms to the normal distribution, this means that the data points are distributed on the normal curve, with most data points concentrated near the mean, while the number of data points farther from the mean gradually decreases. For each data point in the time series, the probability density value under the normal distribution is calculated using the probability density function of the normal distribution, and the probability density value is converted into an outlier probability, that is, the probability that the data point is considered an outlier.

[0065] Exemplarily, an anomaly score is calculated for the reconstructed data, a normal distribution is fitted for the anomaly score, and an outlier probability detection is established based on the normal distribution function, wherein the calculation formula of the anomaly score is:

[0066]

[0067] in, represents the anomaly score of the xth data point calculated after reconstruction; represents all columns of the xth row in the load time series, that is, the xth sequence in the multi-time series; represents the reconstructed x-th target time series; It represents the sum of the squares of the differences between the load time series and the target time series. The higher it is, the greater the possibility that the data point is an outlier.

[0068] In this embodiment, by fitting the abnormal scores of multiple load data points with a normal distribution, for each load data point after the normal distribution fitting, the probability density value of the load data point under the normal distribution is obtained, and the probability density value is converted into an outlier probability. The abnormal data points among the multiple load data points are identified according to the outlier probability, and the outlier probability can be accurately obtained, thereby accurately identifying the abnormal data points among the multiple load data points.

[0069] In an exemplary embodiment, the method further includes: for a time series group composed of multiple target time series, respectively obtaining anomaly scores corresponding to load data points in each target time series in the time series group; under the target dimension, summing up the anomaly scores corresponding to all target time series in the time series group to obtain an anomaly score corresponding to the time series group; and identifying abnormal data points in the time series group based on the anomaly scores corresponding to the time series group.

[0070] Optionally, in the case of multiple time series, it is necessary to sum the anomaly scores calculated for different time series on the same dimension to obtain the anomaly scores corresponding to the time series group. After that, the anomaly scores corresponding to the time series group are fitted with a normal distribution and converted into corresponding outlier probabilities, thereby realizing anomaly detection for multiple time series data.

[0071] In this embodiment, for a time series group composed of multiple target time series, the anomaly scores corresponding to the load data points in each target time series in the time series group are obtained respectively. Under the target dimension, the anomaly scores corresponding to all target time series in the time series group are summed to obtain the anomaly scores corresponding to the time series group. According to the anomaly scores corresponding to the time series group, the abnormal data points in the time series group are identified. This method can be applied to anomaly detection of multiple time series data, thereby improving the comprehensiveness and accuracy of anomaly identification.

[0072] In an exemplary embodiment, the optimized long short-term memory network includes a long short-term memory layer, a fully connected layer, and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0073] Alternatively, in supervised learning, by connecting a fully connected layer to the end of the LSTM, setting the output dimension, and modifying the model's loss function to a binary cross-entropy method, the model outputs a sequence of 0-1 results, which is used to determine whether each data point is an outlier. While retaining the LSTM time series dynamic feature extraction module, a trainable anomaly detection module is added. The optimized long short-term memory network comprises a long short-term memory layer (i.e., LSTM layer), a fully connected layer, and an activation function layer. The LSTM layer extracts the time-dependent features of the load data points, while the fully connected layer learns the decision boundary for anomaly detection through parameter optimization, obtains anomaly decisions, and ultimately outputs the anomaly probability at each time point via the activation function. By incorporating elements of supervised learning, the LSTM model not only captures time series features but also incorporates information from labeled data, further improving its performance and generalization capabilities.

[0074] In this embodiment, by optimizing the long short-term memory network, the scalability and flexibility of the long short-term memory network can be improved, thereby accurately identifying load anomalies in the power system.

[0075] In an exemplary embodiment, the process of identifying an abnormal data point among a plurality of load data points includes: determining that a load data point is an abnormal data point when an outlier probability of the load data point exceeds a first threshold or an abnormal probability of the load data point exceeds a second threshold.

[0076] Optionally, when identifying outliers, in unsupervised learning, a corresponding first threshold is set for the outlier probability. If the outlier probability exceeds the first threshold, it indicates that the data point is an outlier. In supervised learning, a corresponding second threshold is set for the anomaly probability. If the anomaly probability exceeds the second threshold, it indicates that the data point is an outlier.

[0077] In this embodiment, by determining that a load data point is an abnormal data point when the outlier probability of the load data point exceeds a first threshold or the abnormal probability of the load data point exceeds a second threshold, it can be applied to abnormality identification in a variety of scenarios, improve the comprehensiveness of abnormality identification, and thus accurately identify load abnormalities in the power system.

[0078] In an exemplary embodiment, Figure 3As shown, a power system abnormal load point identification system is provided, including: a data acquisition layer, a data processing layer, a model training layer and an application layer.

[0079] The data collection layer is used to deploy automated programs in the power system, obtain current power load, and construct power system load time series data. The automated program can automatically store the collected power load in a CSV file, making it easier for subsequent modules to call.

[0080] The data processing layer cleans and organizes the power load data collected from the power system, checks data integrity, and handles missing and outliers. Detrending is then performed on the time series data to eliminate strong trends while preserving long-range correlations, making the data more suitable for modeling. The data is standardized or normalized to ensure that different features are on the same scale. The historical data is divided into training and test sets and fed into the model for training and subsequent application.

[0081] The model training layer is used to construct the LSTM. During the model training phase, the model is trained using preprocessed data. An appropriate loss function (classification cross entropy) and optimization algorithm are selected. The learning rate, batch size, and number of training rounds are set, and the LSTM is trained using historical long-range correlation sequence data. The trained LSTM is then modified to adapt to unsupervised and supervised scenarios, resulting in the corresponding unsupervised and supervised models.

[0082] The application layer is used to implement anomaly identification and perform corresponding data preprocessing operations on the detection time series, such as detrending and normalization, to achieve time series reconstruction and perform anomaly identification through unsupervised and supervised models.

[0083] In an exemplary embodiment, Figure 4 As shown in the figure, a method for identifying abnormal load points in a power system is provided. Figure 3 Taking this as an example, the following steps are included:

[0084] Obtain power load data of the power system and obtain a load time series based on the power load data.

[0085] The load time series is fitted with a linear function to obtain the first detrended sequence, and the load time series is fitted with a sine function to obtain the second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; the target time series is obtained based on the load time series, the first detrended sequence, and the second detrended sequence.

[0086] The target time series includes load data corresponding to multiple load data points.

[0087] When the long short-term memory network meets the unsupervised learning conditions, the anomaly scores of multiple load data points are obtained through the long short-term memory network, and normal distribution fitting is performed on the anomaly scores of the multiple load data points; for each load data point after normal distribution fitting, the probability density value of the load data point under the normal distribution is obtained, and the probability density value is converted into an outlier probability; and abnormal data points among the multiple load data points are identified according to the outlier probability.

[0088] For a time series group composed of multiple target time series, obtain the anomaly score corresponding to the load data point in each target time series in the time series group respectively; under the target dimension, sum the anomaly scores corresponding to all target time series in the time series group to obtain the anomaly score corresponding to the time series group; and identify the abnormal data points in the time series group based on the anomaly score corresponding to the time series group.

[0089] When the LSTM network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized LSTM network, and abnormal data points among multiple load data points are identified based on the abnormal probability. The optimized LSTM network includes an LSTM layer, a fully connected layer, and an activation function layer; the LSTM layer is used to extract the temporal characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the temporal characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0090] When the outlier probability of the load data point exceeds a first threshold, or the abnormal probability of the load data point exceeds a second threshold, the load data point is determined to be an abnormal data point.

[0091] In this embodiment, the power load data of the power system is obtained, and the load time series is obtained based on the power load data, and the load time series is detrended to obtain a target time series. When the long short-term memory network meets the unsupervised learning conditions, the anomaly scores of multiple load data points are obtained through the long short-term memory network, and abnormal data points among the multiple load data points are identified based on the anomaly scores. When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and abnormal data points among the multiple load data points are identified based on the abnormal probability. This can be applicable to anomaly identification in a variety of scenarios, improve the comprehensiveness of anomaly identification, and thus accurately identify load anomalies in the power system.

[0092] It should be understood that, although the various steps in the flowcharts involved in the above embodiments are displayed in sequence according to the instructions of the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless clearly stated herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps. It is understandable that the various steps in different embodiments can be freely combined as needed, and the various non-contradictory schemes formed by the combination all fall within the scope of protection of this application.

[0093] Based on the same inventive concept, embodiments of the present application further provide a device for identifying abnormal load points in a power system, for implementing the aforementioned method for identifying abnormal load points in a power system. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for identifying abnormal load points in a power system provided below can be found in the aforementioned limitations of the method for identifying abnormal load points in a power system, and will not be further elaborated here.

[0094] In an exemplary embodiment, Figure 5 As shown, a device for identifying abnormal load points in a power system is provided, comprising: a sequence acquisition module 10, a detrending processing module 20, a first identification module 30 and a second identification module 40, wherein:

[0095] The sequence acquisition module 10 is used to acquire power load data of the power system and obtain a load time series based on the power load data.

[0096] The detrending processing module 20 is used to perform detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to a plurality of load data points.

[0097] The first identification module 30 is used to obtain abnormal scores of multiple load data points through the long short-term memory network when the long short-term memory network meets the unsupervised learning conditions, and identify abnormal data points among the multiple load data points according to the abnormal scores.

[0098] The second identification module 40 is used to obtain the abnormal probability of each load data point through the optimized long short-term memory network when the long short-term memory network meets the supervised learning conditions, and identify abnormal data points among multiple load data points based on the abnormal probability.

[0099] In an exemplary embodiment, the detrending processing module 20 is also used to fit the load time series through a linear function to obtain a first detrended sequence, and to fit the load time series through a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; and the target time series is obtained based on the load time series, the first detrended sequence and the second detrended sequence.

[0100] In an exemplary embodiment, the first identification module 30 is also used to fit the abnormal scores of multiple load data points with a normal distribution; for each load data point after the normal distribution is fitted, the probability density value of the load data point under the normal distribution is obtained, and the probability density value is converted into an outlier probability; and abnormal data points among the multiple load data points are identified according to the outlier probability.

[0101] In an exemplary embodiment, the first identification module 30 is also used to obtain the abnormality score corresponding to the load data point in each target time series in the time series group composed of multiple target time series; under the target dimension, the abnormality scores corresponding to all target time series in the time series group are summed to obtain the abnormality score corresponding to the time series group; and according to the abnormality score corresponding to the time series group, the abnormal data points in the time series group are identified.

[0102] In an exemplary embodiment, the optimized long short-term memory network involved in the second recognition module 40 includes a long short-term memory layer, a fully connected layer and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0103] In an exemplary embodiment, the second identification module 40 is further configured to determine that the load data point is an abnormal data point when the outlier probability of the load data point exceeds a first threshold or the abnormal probability of the load data point exceeds a second threshold.

[0104] Each module in the above-mentioned abnormal load point identification device for a power system may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0105] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 6As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication, and the wireless communication can be achieved via Wi-Fi, mobile cellular networks, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for identifying abnormal load points in a power system. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0106] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0107] In an exemplary embodiment, a computer device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: obtaining power load data of a power system, and obtaining a load time series based on the power load data; performing detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to multiple load data points; when a long short-term memory network satisfies unsupervised learning conditions, obtaining anomaly scores of the multiple load data points through the long short-term memory network, and identifying abnormal data points among the multiple load data points based on the anomaly scores; when the long short-term memory network satisfies supervised learning conditions, obtaining anomaly probabilities of each load data point through the optimized long short-term memory network, and identifying abnormal data points among the multiple load data points based on the anomaly probabilities.

[0108] In one embodiment, when a processor executes a computer program, the process of detrending a load time series to obtain a target time series includes: fitting the load time series through a linear function to obtain a first detrended sequence, and fitting the load time series through a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; and obtaining the target time series based on the load time series, the first detrended sequence, and the second detrended sequence.

[0109] In one embodiment, the method of identifying abnormal data points among multiple load data points based on abnormality scores when a processor executes a computer program includes: fitting the abnormality scores of the multiple load data points with a normal distribution; for each load data point after the normal distribution fitting, obtaining a probability density value of the load data point under the normal distribution, and converting the probability density value into an outlier probability; and identifying abnormal data points among the multiple load data points based on the outlier probability.

[0110] In one embodiment, when the processor executes the computer program, the following steps are further implemented: for a time series group composed of multiple target time series, respectively obtaining anomaly scores corresponding to load data points in each target time series in the time series group; under the target dimension, summing the anomaly scores corresponding to all target time series in the time series group to obtain an anomaly score corresponding to the time series group; and identifying abnormal data points in the time series group based on the anomaly scores corresponding to the time series group.

[0111] In one embodiment, the optimized long short-term memory network involved when the processor executes the computer program includes a long short-term memory layer, a fully connected layer, and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0112] In one embodiment, the process of identifying abnormal data points among multiple load data points involved in the execution of a computer program by a processor includes: determining that a load data point is an abnormal data point when the outlier probability of the load data point exceeds a first threshold or the abnormal probability of the load data point exceeds a second threshold.

[0113] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: obtaining power load data of a power system and obtaining a load time series based on the power load data; performing detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to multiple load data points; when a long short-term memory network meets unsupervised learning conditions, obtaining anomaly scores of the multiple load data points through the long short-term memory network, and identifying abnormal data points among the multiple load data points based on the anomaly scores; when the long short-term memory network meets supervised learning conditions, obtaining the abnormal probability of each load data point through the optimized long short-term memory network, and identifying abnormal data points among the multiple load data points based on the abnormal probability.

[0114] In one embodiment, when a computer program is executed by a processor, the process involved in detrending a load time series to obtain a target time series includes: fitting the load time series through a linear function to obtain a first detrended sequence, and fitting the load time series through a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; and obtaining the target time series based on the load time series, the first detrended sequence, and the second detrended sequence.

[0115] In one embodiment, when a computer program is executed by a processor, the method involves identifying abnormal data points among multiple load data points based on abnormality scores, including: fitting a normal distribution to the abnormality scores of the multiple load data points; for each load data point after the normal distribution fitting, obtaining a probability density value of the load data point under the normal distribution, and converting the probability density value into an outlier probability; and identifying abnormal data points among the multiple load data points based on the outlier probability.

[0116] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: for a time series group composed of multiple target time series, respectively obtaining anomaly scores corresponding to load data points in each target time series in the time series group; under the target dimension, summing the anomaly scores corresponding to all target time series in the time series group to obtain an anomaly score corresponding to the time series group; and identifying abnormal data points in the time series group based on the anomaly scores corresponding to the time series group.

[0117] In one embodiment, the optimized long short-term memory network involved when the computer program is executed by a processor includes a long short-term memory layer, a fully connected layer, and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0118] In one embodiment, a computer program executed by a processor involves a process for identifying abnormal data points among a plurality of load data points, comprising determining that a load data point is an abnormal data point when the outlier probability of the load data point exceeds a first threshold or the abnormal probability of the load data point exceeds a second threshold.

[0119] In one embodiment, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the following steps: obtaining power load data of a power system and obtaining a load time series based on the power load data; performing detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to a plurality of load data points; obtaining anomaly scores of the plurality of load data points through the long short-term memory network when a long short-term memory network satisfies unsupervised learning conditions, and identifying abnormal data points among the plurality of load data points based on the anomaly scores; obtaining an abnormal probability of each load data point through an optimized long short-term memory network when a long short-term memory network satisfies supervised learning conditions, and identifying abnormal data points among the plurality of load data points based on the anomaly probabilities.

[0120] In one embodiment, when a computer program is executed by a processor, the process involved in detrending a load time series to obtain a target time series includes: fitting the load time series through a linear function to obtain a first detrended sequence, and fitting the load time series through a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; and obtaining the target time series based on the load time series, the first detrended sequence, and the second detrended sequence.

[0121] In one embodiment, when a computer program is executed by a processor, the method involves identifying abnormal data points among multiple load data points based on abnormality scores, including: fitting a normal distribution to the abnormality scores of the multiple load data points; for each load data point after the normal distribution fitting, obtaining a probability density value of the load data point under the normal distribution, and converting the probability density value into an outlier probability; and identifying abnormal data points among the multiple load data points based on the outlier probability.

[0122] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: for a time series group composed of multiple target time series, respectively obtaining anomaly scores corresponding to load data points in each target time series in the time series group; under the target dimension, summing the anomaly scores corresponding to all target time series in the time series group to obtain an anomaly score corresponding to the time series group; and identifying abnormal data points in the time series group based on the anomaly scores corresponding to the time series group.

[0123] In one embodiment, the optimized long short-term memory network involved when the computer program is executed by a processor includes a long short-term memory layer, a fully connected layer, and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; and the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

[0124] In one embodiment, a computer program executed by a processor involves a process for identifying abnormal data points among a plurality of load data points, comprising determining that a load data point is an abnormal data point when the outlier probability of the load data point exceeds a first threshold or the abnormal probability of the load data point exceeds a second threshold.

[0125] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0126] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0127] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for identifying abnormal load points in a power system, characterized in that: The method comprises: Acquiring power load data of the power system, and acquiring a load time series based on the power load data; Detrending the load time series to obtain a target time series; the target time series includes load data corresponding to a plurality of load data points; When the long short-term memory network satisfies an unsupervised learning condition, obtaining abnormal scores of the multiple load data points through the long short-term memory network, and identifying abnormal data points among the multiple load data points according to the abnormal scores; When the long short-term memory network meets the supervised learning conditions, the abnormal probability of each load data point is obtained through the optimized long short-term memory network, and the abnormal data points among the multiple load data points are identified according to the abnormal probability.

2. The method according to claim 1, characterized in that The detrending process is performed on the load time series to obtain a target time series, including: Fitting the load time series by a linear function to obtain a first detrended sequence, and fitting the load time series by a sine function to obtain a second detrended sequence; the first detrended sequence is used to characterize the linear trend of the load time series; the second detrended sequence is used to characterize the periodic trend of the load time series; A target time series is obtained according to the load time series, the first detrended sequence, and the second detrended sequence.

3. The method according to claim 1, characterized in that The identifying an abnormal data point among the plurality of load data points according to the abnormality score comprises: performing normal distribution fitting on the abnormal scores of the multiple load data points; For each load data point after normal distribution fitting, obtain the probability density value of the load data point under the normal distribution, and convert the probability density value into an outlier probability; Abnormal data points among the plurality of load data points are identified based on the outlier probability.

4. The method according to claim 3, characterized in that The method further comprises: For a time series group composed of a plurality of target time series, respectively obtaining an anomaly score corresponding to a load data point in each target time series in the time series group; Under the target dimension, sum the anomaly scores corresponding to all target time series in the time series group to obtain the anomaly score corresponding to the time series group; Abnormal data points in the time series group are identified according to the anomaly scores corresponding to the time series group.

5. The method according to claim 1, wherein The optimized long short-term memory network includes a long short-term memory layer, a fully connected layer and an activation function layer; the long short-term memory layer is used to extract the time characteristics of the load data points; the fully connected layer is used to obtain abnormal decisions based on the time characteristics; the activation function layer is used to obtain the abnormal probability of each load data point based on the abnormal decision.

6. The method according to claim 3, characterized in that The process of identifying abnormal data points among the plurality of load data points comprises: When the outlier probability of the load data point exceeds a first threshold, or the abnormal probability of the load data point exceeds a second threshold, the load data point is determined to be an abnormal data point.

7. A device for identifying abnormal load points in a power system, characterized in that: The device comprises: A sequence acquisition module is used to acquire power load data of the power system and obtain a load time series based on the power load data; a detrending processing module, configured to perform detrending processing on the load time series to obtain a target time series; the target time series includes load data corresponding to a plurality of load data points; a first identification module, configured to obtain, by using the long short-term memory network, anomaly scores of the plurality of load data points when the long short-term memory network satisfies an unsupervised learning condition, and identify abnormal data points among the plurality of load data points according to the anomaly scores; The second identification module is used to obtain the abnormal probability of each load data point through the optimized long short-term memory network when the long short-term memory network meets the supervised learning conditions, and identify the abnormal data points among the multiple load data points according to the abnormal probability.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images