A method for monitoring and early warning of substation ground current micro-power consumption in synchronization

Abstract

The application discloses a kind of for substation ground current micro-power consumption synchronous monitoring early warning method, it is related to substation monitoring technical field, the method includes: obtaining ground current dataset;Dynamic time warping is carried out to ground current dataset, and time sequence synchronization dataset is obtained;Time series analysis is carried out based on time sequence synchronization dataset, and multiple ground current characteristics are determined;Current fluctuation learning is carried out by traversing time sequence synchronization dataset, and ground current prediction dataset is generated;The grounding network of substation is monitored dynamically, and multiple feedback response signals are generated to the intelligent triggering alarm of substation.Solve the technical problems that the ground current data of multiple grounding points cannot be collected in real time, real-time abnormal fluctuation cannot be accurately captured in the prior art, which leads to the inability to identify potential risks of ground current in advance and effectively, resulting in insufficient real-time and accuracy of early warning, realize the low-power continuous monitoring of ground current, and achieve the technical effect of improving the real-time and accuracy of monitoring and early warning.

Description

Technical Field

[0001] This application relates to the field of substation monitoring technology, and in particular to a synchronous monitoring and early warning method for low power consumption of substation ground current. Background Technology

[0002] With the continuous development and increasing intelligence of power systems, substations, as key nodes in the power system, are crucial to the reliability of the entire power grid due to their safe and stable operation. The substation grounding network, as an important component of the safety protection system, is responsible for safely guiding fault currents or overcurrents generated by equipment and systems to the ground in the event of power system faults or anomalies, protecting equipment and personnel. However, in actual operation, the grounding network may be affected by factors such as equipment aging, environmental changes, fault currents, and lightning strikes, leading to fluctuations and anomalies in ground currents. Especially in large-scale power systems, the characteristics of ground currents at different grounding points vary significantly and change frequently, potentially causing equipment failures or even power grid accidents, thus placing higher demands on the monitoring of the grounding network's operation. Traditional ground current monitoring mainly relies on periodic sampling to acquire ground current data from various grounding points to analyze current fluctuations. This method suffers from time lag in data acquisition and cannot simultaneously monitor data changes at multiple grounding points, making it difficult to accurately capture real-time abnormal fluctuation characteristics and effectively identify dynamic patterns and potential trends in ground current changes. Consequently, ground current cannot be monitored and warned of in a timely and effective manner, potentially leading to grounding system overload, equipment damage, or even endangering personnel safety.

[0003] Current technologies for synchronous monitoring and early warning of substation ground current have several limitations. They cannot collect ground current data from multiple grounding points in real time, and it is difficult to accurately capture real-time abnormal fluctuations. Consequently, they cannot identify potential risks of ground current in advance, resulting in insufficient real-time performance and accuracy of early warnings. Summary of the Invention

[0004] This application provides a synchronous monitoring and early warning method for low-power ground current in substations, which solves the technical problems in the prior art, such as the inability to collect ground current data from multiple grounding points in real time, the difficulty in accurately capturing real-time abnormal fluctuations, and the inability to identify potential risks of ground current in advance, resulting in insufficient real-time and accuracy of early warning. This method achieves low-power continuous monitoring of ground current, thereby improving the real-time performance and accuracy of monitoring and early warning.

[0005] This application provides a synchronous monitoring and early warning method for low-power ground current in substations, comprising: sensing multiple equipment grounding points in the substation through the substation's grounding network to obtain a ground current dataset; traversing the multiple equipment grounding points to determine the acquisition sequence, and dynamically time-warping the ground current dataset according to the acquisition sequence to obtain a time-series synchronous dataset; performing time series analysis based on the time-series synchronous dataset, extracting features based on the analysis results, and determining multiple ground current features; traversing the time-series synchronous dataset according to the multiple ground current features to perform current fluctuation learning, and generating a ground current prediction dataset based on the learning results; and dynamically monitoring the substation's grounding network based on the ground current prediction dataset, generating multiple feedback response signals to intelligently trigger alarms for the substation.

[0006] In a possible implementation, multiple equipment grounding points in the substation are sensed through the substation's grounding network to obtain a ground current dataset. The following processing is also performed: a coverage traversal of the substation based on the grounding network is performed to obtain multiple acquisition nodes; current is captured at the substation based on the multiple acquisition nodes, and the multiple equipment grounding points are determined based on the current capture results; multiple sensing devices are activated and a acquisition frequency is set according to the substation's operating status information; the multiple sensing devices are used to traverse and sense the multiple equipment grounding points according to the acquisition frequency to obtain multiple ground current data signals, each containing a timestamp; the multiple ground current data signals are integrated to determine the ground current dataset.

[0007] In a possible implementation, the acquisition timing is determined by traversing the multiple device grounding points. The ground current dataset is then dynamically time-warped according to the acquisition timing to obtain a time-synchronized dataset. The following processing is also performed: the acquisition timing is determined by traversing the multiple device grounding points and combining the timestamp of each ground current data signal; the acquisition timing is used as a timing reference for node selection to determine a timing baseline; the multiple ground current data signals are amplified to obtain multiple ground current enhancement signals; the multiple ground current enhancement signals are dynamically time-warped based on the timing baseline to determine the signal time step; the ground current dataset is interpolated and padded according to the signal time step to generate multiple time windows; the multiple time windows are validated for consistency; and the multiple time windows are correlated and integrated based on the validation results to output the time-synchronized dataset.

[0008] In a possible implementation, dynamic time warping is performed on the plurality of ground current enhancement signals based on the time-series reference to determine the signal time step. The following processing is also performed: similarity calculation is performed on the plurality of ground current enhancement signals based on the time-series reference to obtain multiple similarity coefficients; a two-dimensional matrix is ​​constructed based on the time-series reference, and Euclidean distance is calculated on the plurality of ground current enhancement signals according to the multiple similarity coefficients to generate multiple distance data; the multiple distance data are mapped to the two-dimensional matrix for path optimization to determine a time-aligned path; and the plurality of ground current enhancement signals are time-aligned based on the time-aligned path to generate the signal time step.

[0009] In a possible implementation, current fluctuation learning is performed by traversing the time-series synchronous dataset according to the multiple ground current features, and a ground current prediction dataset is generated based on the learning results. The following processing is also performed: sliding windows are obtained by combining the multiple time windows with the multiple ground current features; multiple current dynamic fluctuation data are obtained by traversing and capturing the multiple sliding windows; a long short-term memory neural network is used to learn the multiple current dynamic fluctuation data to generate learning results; trend analysis is performed based on the learning results and the time-series synchronous dataset to draw a current fluctuation curve; training is performed based on the current fluctuation curve and the learning results to obtain training results; prediction is performed based on the training results and the multiple ground current features according to the time-series benchmark to generate a ground current prediction data sequence; and the ground current prediction data sequence is added to the ground current prediction dataset.

[0010] In a possible implementation, the substation's grounding network is dynamically monitored based on the ground current prediction dataset, generating multiple feedback response signals to intelligently trigger alarms for the substation. The following processing is also performed: retrieving multiple operating condition information sets; initiating operational simulation analysis of the grounding network based on the multiple operating condition information sets to obtain multiple ground current operational simulation data sets; performing a traversal matching process between the multiple ground current operational simulation data sets and the ground current prediction dataset to generate multiple simulation-prediction data pairs; dynamically comparing the multiple simulation-prediction data pairs and generating multiple feedback response signals based on the comparison results; traversing the multiple feedback response signals and combining them with the multiple simulation-prediction data pairs to perform current change analysis and generate multiple anomaly tags; identifying the multiple feedback response signals based on the multiple anomaly tags to determine multiple anomaly signals; and intelligently triggering alarms based on the multiple anomaly signals.

[0011] In a possible implementation, the process involves iterating through the multiple feedback response signals and combining them with the multiple simulation-prediction data pairs to perform current change analysis, generating multiple anomaly tags, and further performing the following steps: performing current feedback analysis based on the multiple feedback response signals to determine multiple current change characteristics; setting a current change deviation threshold according to the multiple current change characteristics; calculating the error of the multiple simulation-prediction data pairs to generate multiple current change error values; comparing the multiple current change error values ​​with the current change deviation threshold to determine whether the multiple current change error values ​​are greater than or equal to the current change deviation threshold; and if the multiple current change error values ​​are greater than or equal to the current change deviation threshold, then identifying the simulation-prediction data pairs corresponding to the current change error values ​​greater than or equal to the current change deviation threshold to generate the multiple anomaly tags.

[0012] In a possible implementation, the multiple feedback response signals are identified based on the multiple anomaly tags to determine multiple anomaly signals. Intelligent alarm triggering is then performed based on these multiple anomaly signals. The following processes are also performed: using the multiple anomaly tags as indexes, the multiple feedback response signals are traversed and matched to determine multiple anomaly signals; an impact assessment is performed based on the multiple anomaly signals, and multiple impact levels are generated based on the assessment results; the multiple anomaly signals are categorized according to the multiple impact levels to generate signal categorization results; multiple alarm triggering conditions are set based on the signal categorization results, and intelligent alarm triggering is performed based on the multiple alarm triggering conditions combined with the multiple anomaly signals.

[0013] This application proposes a method for low-power synchronous monitoring and early warning of substation ground current. The method involves obtaining a ground current dataset; dynamically warping the ground current dataset to obtain a time-series synchronous dataset; performing time series analysis on the time-series synchronous dataset to determine multiple ground current characteristics; traversing the time-series synchronous dataset to learn current fluctuations and generate a ground current prediction dataset; and dynamically monitoring the substation's grounding network to generate multiple feedback response signals for intelligent triggering of alarms. This method solves the technical problems in existing technologies, such as the inability to collect ground current data from multiple grounding points in real time and the difficulty in accurately capturing real-time abnormal fluctuations, leading to the inability to effectively identify potential risks of ground current in advance, resulting in insufficient real-time and accurate early warning. It achieves low-power continuous monitoring of ground current, thus improving the real-time performance and accuracy of monitoring and early warning. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 A flowchart illustrating a synchronous monitoring and early warning method for low power consumption of substation ground current provided in an embodiment of this application;

[0016] Figure 2 This is a flowchart illustrating the process of obtaining a timing synchronization dataset in a synchronous monitoring and early warning method for low power consumption of substation ground current provided in an embodiment of this application. Detailed Implementation

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0020] This application provides a synchronous monitoring and early warning method for low power consumption of substation ground current, such as... Figure 1 As shown, the method includes:

[0021] Step S100: Sensing multiple equipment grounding points in the substation through the substation's grounding network to obtain a ground current dataset.

[0022] Preferably, the grounding network typically covers multiple key devices in the substation, including transformers, switchgear, busbars, control cabinets, etc. These devices each have their own grounding points, responsible for guiding excess current to the ground in the event of a fault or abnormal situation to ensure equipment safety. Specifically, current transformers, Hall effect sensors, or low-power measurement devices are installed at the grounding points of each key device to monitor the magnitude and changes of the current at each device's grounding point in real time. These current sensors are connected to the substation grounding network and collect ground current data at the grounding points by detecting the current flow in the grounding system. This includes the current value of each grounding point and its changes at different times, such as instantaneous current values, periodic current fluctuations, and abnormal current peaks. This forms a ground current dataset containing multiple time points and multiple grounding points. The ground current dataset includes the ground current record of each device's grounding point, that is, it contains the ground current distribution characteristics of the substation equipment under normal and abnormal conditions.

[0023] In one possible implementation, step S100 further includes step S110, performing a coverage traversal of the substation based on the grounding network to obtain multiple acquisition nodes of the substation; step S120, capturing current in the substation according to the multiple acquisition nodes, and determining the multiple device grounding points based on the current capture results; step S130, activating multiple sensing devices and setting the acquisition frequency according to the substation operating status information; step S140, performing traversal sensing on the multiple device grounding points according to the acquisition frequency based on the multiple sensing devices to obtain multiple ground current data signals, each ground current data signal containing a timestamp; and step S150, integrating the multiple ground current data signals to determine the ground current dataset.

[0024] Preferably, the substation's grounding network is traversed to determine key locations within it, resulting in multiple data acquisition nodes. These nodes are typically located at critical grounding points in the grounding system, including transformer grounding points, busbar grounding points, and switchgear grounding points. At each acquisition node, sensing devices (such as current transformers or Hall effect sensors) are used to capture current signals and record current data in real time. The current capture results are then used to further confirm the actual location and connection status of each device's grounding point. Finally, based on the substation's operating status information (such as load conditions and equipment operating frequencies), the sensing devices at each acquisition node are automatically activated, and the acquisition frequency is set. For example, the sampling frequency is increased during extreme conditions such as thunderstorms, high loads, or equipment malfunctions to ensure that all critical current fluctuation information is captured; while the sampling frequency is reduced under normal conditions to save energy (e.g., once per second); based on the set sampling frequency of each acquisition node, the ground current signal of each grounding point is collected using sensing devices, and each signal contains a timestamp to record the specific time of sampling; finally, the ground current data signals of each acquisition node are integrated, including processing such as timestamp alignment, signal denoising, and data correction, to form a ground current dataset, providing real-time and accurate monitoring data for the monitoring and fault diagnosis of substation grounding current.

[0025] Step S200: Traverse the multiple device grounding points to determine the acquisition timing sequence, and perform dynamic time warping on the ground current dataset according to the acquisition timing sequence to obtain a timing synchronization dataset.

[0026] Preferably, the substation has multiple equipment grounding points (such as transformer grounding points, busbar grounding points, switchgear grounding points, etc.), each with an independent ground current sensor. Specifically, the multiple equipment grounding points within the substation are traversed to determine the acquisition sequence, i.e., assigning an acquisition time sequence to each grounding point, i.e., setting a polling mechanism. This acquisition sequence can be optimized based on the physical location of the sensor, signal transmission delay, or other factors, making the acquisition process both efficient and meeting the system's time synchronization requirements. Then, the acquired ground current dataset is dynamically time-warped according to the acquisition sequence. Specifically, by dynamically adjusting the time axis of each time series, the ground current data from different sampling times can be aligned under the same time reference, i.e., adjusting the time stamp of the sampling points to keep the data of each grounding point synchronized in time. Dynamic time warping (Dynamic Time) is a key aspect of this process. Time-series data (DTW) is used to align time series data and is particularly suitable for handling time deviations between different data acquisition points. Since the sampling times of different grounding points are slightly different, directly comparing their time series data may result in time misalignment. Finally, a time-series synchronization dataset is generated, which contains the ground current values ​​of each substation grounding point at the same time. This dataset is used to analyze the ground current relationship between equipment, identify synchronization fluctuation patterns, and improve the accuracy of anomaly detection.

[0027] In one possible implementation, such as Figure 2 As shown, step S200 further includes step S210, traversing the multiple device grounding points and performing time-series processing based on the timestamp of each ground current data signal to determine the acquisition timing; step S220, using the acquisition timing as a timing reference for node selection to determine the timing benchmark; step S230, amplifying the multiple ground current data signals to obtain multiple ground current enhancement signals; step S240, performing dynamic time warping on the multiple ground current enhancement signals based on the timing benchmark to determine the signal time step; step S250, interpolating and padding the ground current dataset according to the signal time step to generate multiple time windows; step S260, performing consistency verification on the multiple time windows, and associating and integrating the multiple time windows according to the verification results to output the timing synchronization dataset.

[0028] Preferably, the grounding points of each device in the substation are traversed to acquire ground current data signals from each grounding point. Each signal carries a timestamp. The acquired ground current data signals are sorted according to the timestamps to generate an acquisition timing sequence, which reflects the temporal order of the ground current data. This sequence is then used as a timing reference. A key node or a set of nodes is selected, such as the average sampling time of each device or the earliest occurrence time of the data, as the timing reference. This reference node's acquisition timing sequence is used as the timing benchmark and serves as the standard time axis for subsequent dynamic time warping, ensuring that all data points are aligned within a unified time frame. Furthermore, multiple ground current data signals are amplified to enhance signal amplitude and improve observability, resulting in multiple enhanced ground current signals. This ensures that the signal amplitude and frequency are on the same order of magnitude, facilitating subsequent timing alignment and analysis.

[0029] Preferably, when the signal time step is determined, Dynamic Time Warping (DTW) is used to align the enhanced signals. This involves adjusting the signal time step to align the enhanced signals from each grounding point to the same timing reference, thus determining the signal time step. This signal time step represents the time interval of each time step, ensuring all data points are on the same time scale. In each enhanced signal, there may be some missing data points. According to the determined signal time step, the ground current dataset is interpolated to fill in the missing data points, ensuring each signal sequence has a continuous time series. The interpolated data is divided into multiple time windows according to fixed time intervals, each representing a different time period of the ground current data. Finally, a consistency check is performed on the generated multiple time windows to verify the rationality and consistency of the data within each window, ensuring there is no bias or erroneous data. Based on the consistency check results, the data from each time window are correlated and integrated to ensure consistency in time and characteristics between time windows, thereby outputting a time-synchronized dataset that integrates the ground current data signals from all device grounding points, possessing a unified time step, time reference, and complete time windows.

[0030] In one possible implementation, step S240 further includes step S241, calculating the similarity of the plurality of ground current enhancement signals based on the timing reference to obtain a plurality of similarity coefficients; step S242, constructing a two-dimensional matrix according to the timing reference, calculating the Euclidean distance of the plurality of ground current enhancement signals according to the plurality of similarity coefficients to generate a plurality of distance data; step S243, mapping the plurality of distance data to the two-dimensional matrix for path optimization to determine a timing alignment path; and step S244, aligning the plurality of ground current enhancement signals in time based on the timing alignment path to generate the signal time step.

[0031] Preferably, similarity calculation and Euclidean distance calculation are used to determine the optimal alignment path for each signal to achieve time synchronization. Specifically, based on a set time series reference, the similarity of multiple ground current enhancement signals is calculated, such as using cosine similarity, Pearson correlation coefficient, or dynamic time warping similarity, to quantify the similarity between each pair of signals. The similarity calculation result for each pair of signals is called the similarity coefficient. The higher the similarity coefficient, the more similar the change patterns of the two signals are. A two-dimensional matrix is ​​constructed based on the time series reference, where each element represents the similarity of a pair of ground current enhancement signals at a specific time point. According to the similarity coefficient, Euclidean distance is calculated for each time point of each pair of ground current enhancement signals to quantify the degree of difference between the two signals at a certain time point. The smaller the distance, the more similar the change patterns of the two signals are. The higher the similarity at a given time point, the more distance data are generated. These multiple distance data are then mapped onto a two-dimensional matrix. Dynamic programming or dynamic time warping (DTW) is used to find an optimal path in the matrix that minimizes the cumulative distance along the path. This path serves as the timing alignment path, representing the best alignment of multiple signals in time, ensuring that the signals are synchronized on the time axis. Finally, based on the determined timing alignment path, multiple ground current enhancement signals are adjusted in time to align them on the time axis. This includes adjusting the time steps of each signal to ensure they remain consistent at every moment. The resulting signal time step reflects the alignment interval at each time point, reducing errors caused by differences in acquisition timing. This provides a high-precision timing reference for timing analysis and anomaly detection.

[0032] Step S300: Perform time series analysis based on the time-series synchronization dataset, extract features based on the analysis results, and determine multiple ground current features.

[0033] Preferably, using the time-synchronized ground current dataset, time series analysis techniques are employed to identify and extract key features reflecting ground current characteristics. Specifically, time series analysis determines the long-term trend of ground current variation, reflecting the overall changes in the grounding system over a period of time (e.g., a gradually increasing current may indicate an increase in grounding resistance or equipment aging). Autocorrelation and periodogram methods are used to detect potential periodic components in the ground current signal, obtaining periodic features and identifying abrupt changes in the ground current signal, such as instantaneous current peaks or abnormal fluctuations, which may indicate instantaneous faults or abnormal conditions (e.g., short circuits, changes in ground resistance). The stability of the ground current data over time is checked; if the signal exhibits significant non-stationarity, further analysis is performed. The analysis identifies potential sources of anomalies. Based on time series analysis, multiple features are extracted from the ground current data, containing core characteristic information of the ground current signal. This information can be used to monitor the status of the grounding network in real time. For example, the mean and standard deviation are used to describe the central trend and fluctuation degree of the ground current. The maximum and minimum values ​​and their intervals in the ground current are extracted to reflect extreme current fluctuations, which helps to identify sudden anomalies. The intensity of the periodic component is quantified by methods such as Fourier transform. The periodic intensity reflects the repetition pattern of the ground current, which may differ under normal and abnormal operating conditions. The autocorrelation coefficient is used to measure the correlation of the ground current at different time lags. The time change rate of the ground current signal is calculated to identify the frequency and degree of fluctuations or abrupt changes.

[0034] Step S400: Traverse the time-series synchronization dataset according to the multiple ground current features to learn about current fluctuations, and generate a ground current prediction dataset based on the learning results.

[0035] Preferably, the ground current characteristics in the time-series synchronous dataset are analyzed and modeled to learn the fluctuation patterns and variation laws of the ground current, thereby predicting the future trend of the ground current. Specifically, the time-series synchronous dataset is traversed based on multiple ground current characteristics, that is, the ground current time-series data of each substation equipment grounding point is analyzed to capture the historical fluctuation of the ground current at each grounding point and discover its variation patterns. Then, machine learning models or deep learning models are used to learn the current fluctuations. For example, a fluctuation prediction model is built based on time-series models such as ARIMA, LSTM (Long Short-Term Memory Network), smoothing filtering, or regression analysis. The aim is to extract the inherent laws of ground current fluctuations by analyzing the variation patterns of the ground current characteristics at each grounding point over time, capture its time series characteristics, and predict the ground current characteristic values ​​for future times based on the fluctuation patterns in historical data. For example, the LSTM model can predict the trend of ground current changes in the next few minutes or hours. The predicted ground current characteristic values ​​are combined to form a ground current prediction dataset, which contains the predicted ground current values ​​and fluctuations at each grounding point at future times, as a possible distribution of future ground currents.

[0036] In one possible implementation, step S400 further includes step S410, sliding multiple time windows based on the multiple ground current features to obtain multiple sliding windows; step S420, traversing and capturing multiple current dynamic fluctuation data according to the multiple sliding windows to obtain multiple current dynamic fluctuation data; step S430, using a long short-term memory neural network to learn the multiple current dynamic fluctuation data to generate learning results; step S440, performing trend analysis according to the learning results and the time-series synchronization dataset to draw a current fluctuation curve; step S450, training based on the current fluctuation curve and the learning results to obtain training results, and predicting according to the time-series benchmark based on the training results and the multiple ground current features to generate a ground current prediction data sequence; step S460, adding the ground current prediction data sequence to the ground current prediction dataset.

[0037] Preferably, the ground current data is divided into multiple time windows along a time axis. Each window contains ground current data for a certain period. A sliding window technique is used to slide the data along the time axis with a fixed step size, generating multiple sliding windows containing different time periods. The dynamic fluctuation data of the ground current is captured by traversing these sliding windows, and ground current features (such as peak value, rate of change, periodicity, etc.) are extracted. The dynamic fluctuation data reflects the instantaneous changes of the ground current within each sliding window. Then, a prediction model is built based on a Long Short-Term Memory (LSTM) neural network. The captured dynamic fluctuation data is input into the prediction model for training, learning the inherent laws and time-dependent characteristics of ground current fluctuations. The learning results are output, including the model's understanding and prediction ability of future ground current fluctuation trends. LSTM is a deep learning model suitable for time series analysis, capable of capturing… The study investigates the dependencies between long-term and short-term data. Combining the learning results with a time-series synchronized dataset, it performs trend analysis on ground current data. Specifically, by analyzing the changing trend of ground current over time, it identifies its main patterns, such as upward trends and periodic fluctuations. The trend analysis results are then presented as a curve graph, creating a current fluctuation curve. Based on the plotted current fluctuation curve and the learning results, the prediction model is trained to better capture the fluctuation characteristics of ground current, improving its prediction accuracy and sensitivity to abnormal fluctuations. Finally, combining the training results and multiple ground current features, and according to a time-series benchmark, the study predicts future ground current fluctuations, generating a predicted data sequence that reflects the ground current fluctuations at various grounding points in the future. This generated ground current prediction data sequence is then added to the ground current prediction dataset.

[0038] Step S500: Dynamically monitor the substation's grounding network based on the ground current prediction dataset, and generate multiple feedback response signals to intelligently trigger alarms for the substation.

[0039] Preferably, the system utilizes the future ground current trend provided by the predicted dataset to dynamically monitor the operating status of the grounding network in real time, generating feedback signals for abnormal situations and triggering intelligent alarms for timely response. Specifically, the expected ground current value in the predicted dataset is compared with the actual collected ground current data to detect any anomalies. A dynamic threshold is set based on the predicted data; if the actual data exceeds the predicted range or reaches a certain threshold, it may indicate an anomaly or potential fault in the grounding network. Dynamic monitoring refers to automatically adjusting the focus and frequency of monitoring based on changes in the predicted data. For example, when the predicted data indicates an increased fluctuation in the ground current at a certain grounding point, the system can increase monitoring at that point. The system uses frequency to capture more detailed real-time data, generating multiple feedback response signals, including warning signals, risk alert signals, and abnormal alarm signals. When the deviation between actual and predicted data reaches a set threshold, a corresponding feedback signal is generated and categorized into different levels (e.g., early warning, alert, emergency alarm). Simultaneously, it automatically sends real-time feedback through scheduling sensors and intelligently triggers alarms. That is, intelligent alarms can be categorized by type and priority. For example, minor deviations or fluctuations trigger low-level alarms to alert maintenance personnel; significant deviations or high-risk fluctuations trigger high-level alarms, indicating potential fault risks requiring immediate inspection and handling. Intelligent alarms can be categorized according to the priority and triggering conditions of different response signals for accurate alarm delivery, optimizing the response speed and processing efficiency of maintenance personnel, thereby improving the safety and stability of the substation grounding network.

[0040] In one possible implementation, step S500 further includes step S510: retrieving multiple operating condition information, starting the grounding network for operational simulation analysis based on the multiple operating condition information, and obtaining multiple ground current operational simulation data; step S520: performing traversal matching between the multiple ground current operational simulation data and the ground current prediction dataset to generate multiple simulation-prediction data pairs; step S530: dynamically comparing the multiple simulation-prediction data pairs, and generating multiple feedback response signals based on the comparison results; step S540: traversing the multiple feedback response signals and combining them with the multiple simulation-prediction data pairs to perform current change analysis, and generating multiple anomaly tags; step S550: identifying the multiple feedback response signals based on the multiple anomaly tags, determining multiple anomaly signals, and intelligently triggering alarms based on the multiple anomaly signals.

[0041] Preferably, by comparing and analyzing simulations and predictions, abnormal ground current conditions are identified and alarms are triggered. Specifically, operating condition information (various conditions of the substation under different operating states, such as load changes, equipment operation, ambient temperature, etc.) is retrieved, and the grounding network simulation analysis is initiated based on the operating condition information. That is, by simulating the behavior of the grounding network under different operating conditions, the ground current distribution under different conditions is obtained, and multiple ground current operation simulation data are generated, reflecting the theoretical ground current characteristics under various operating conditions. The simulation data is matched with the ground current prediction dataset one by one to find the correspondence between each simulation data point and the prediction data point, forming a simulation-prediction data pair. Each simulation-prediction data pair contains simulation data and prediction data at the same time and the same grounding point, which are used to compare the differences between the actual operating state and the predicted state.

[0042] Preferably, multiple simulation-prediction data pairs are dynamically compared, such as comparing current magnitude, fluctuation amplitude, and trend, to generate feedback response signals. Each data pair corresponds to a signal, regardless of the magnitude of the difference or whether the data pair is abnormal or normal. Then, all feedback response signals are iterated and combined with the corresponding simulation-prediction data pairs to analyze the actual changes in ground current, including trend analysis, abnormal amplitude assessment, and periodic change identification. This identifies which changes exceed the normal range. Based on the current change analysis results, multiple abnormal labels are generated to identify specific abnormal types and characteristics (e.g., excessive fluctuation, sudden increase, continuous exceedance, etc.). Finally, the feedback signals are labeled with information such as the cause of the abnormality, location of occurrence, and degree of abnormality to distinguish different types of abnormalities, thus forming multiple abnormal signals that reflect the abnormal state in the grounding network. Intelligent alarms are triggered based on these abnormal signals.

[0043] In one possible implementation, step S540 further includes step S541, performing current feedback analysis based on the plurality of feedback response signals to determine a plurality of current change characteristics; step S542, setting a current change deviation threshold according to the plurality of current change characteristics, performing error calculation on the plurality of simulation-prediction data pairs to generate a plurality of current change error values; step S543, comparing the plurality of current change error values ​​with the current change deviation threshold to determine whether the plurality of current change error values ​​are greater than or equal to the current change deviation threshold; step S544, if the plurality of current change error values ​​are greater than or equal to the current change deviation threshold, then identifying the simulation-prediction data pairs corresponding to the current change error values ​​greater than or equal to the current change deviation threshold to generate the plurality of anomaly tags.

[0044] Preferably, the actual feedback of the ground current is analyzed based on multiple feedback response signals to identify the current change patterns and trends, and to determine multiple current change characteristics, which may include the average rate of change, fluctuation amplitude, abnormal peak values, periodic fluctuations, etc., reflecting the change law of the ground current under different conditions. Based on the determined current change characteristics, multiple corresponding deviation thresholds are set as benchmarks for judging current change anomalies, reflecting the fluctuation range of the ground current under normal conditions. The deviation thresholds are used for subsequent error calculation and comparison; the stricter the deviation threshold, the higher the sensitivity to abnormal changes. The actual current data of each simulation-prediction data pair is compared with the predicted current data, and the error between them is calculated, called the current change error value. Each error value represents the degree of deviation between a simulation-prediction data pair. The larger the error value, the more significant the difference between the simulation and prediction data, and the more likely there is an anomaly. Then, the generated current change error values ​​are compared one by one with the set current change deviation threshold. That is, the error values ​​are compared to see if they exceed the deviation threshold to identify abnormal changes. If the current change error value is greater than or equal to the current change deviation threshold, it means that the error value has reached or exceeded the allowable fluctuation range, which may indicate an abnormal current situation. For simulation-prediction data pairs with error values ​​exceeding the deviation threshold, they are marked as abnormal data. Based on the marking results, multiple abnormal labels are generated for current change error values ​​that exceed the deviation threshold, such as "mild anomaly", "moderate anomaly", "severe anomaly", etc.

[0045] In one possible implementation, step S550 further includes step S551, using the multiple abnormal tags as indexes to traverse and match the multiple feedback response signals to determine multiple abnormal signals; step S552, performing an impact assessment based on the multiple abnormal signals and generating multiple impact levels based on the assessment results; step S553, sorting the multiple abnormal signals according to the multiple impact levels to generate signal sorting results; and step S554, setting multiple alarm triggering conditions based on the signal sorting results, and intelligently triggering alarms based on the multiple alarm triggering conditions combined with the multiple abnormal signals.

[0046] Preferably, all feedback response signals are traversed, matched against anomaly labels, and each feedback signal is identified as containing an abnormal condition. All feedback signals matching the anomaly label are identified as anomalous signals. Multiple anomalous signals are identified, representing potential anomalies in the substation. Then, an impact assessment is performed on each anomalous signal to quantify its potential impact on the grounding network and the overall system. This assessment includes evaluating factors such as current fluctuation amplitude, duration, and frequency of anomaly occurrence. Based on the impact assessment results, each anomalous signal is assigned an impact level, such as multiple levels (e.g., low, medium, high), to indicate the severity and urgency of the anomalous signal. A higher impact level indicates a greater potential threat to the system. All abnormal signals are categorized according to their impact level, separating signals with varying degrees of impact to generate a signal categorization result. For example, severe abnormal signals are assigned to the highest-level group, while minor abnormal signals are assigned to lower-level groups. Finally, based on the categorization results of different impact levels, corresponding alarm trigger conditions are set for each group of abnormal signals. These alarm trigger conditions include response level, alarm method, and processing priority. Based on different trigger conditions, alarm signals of different urgency levels are issued, and intelligent alarm decisions are made automatically. This enables more accurate identification and response to anomalies of different levels, reduces false alarms and missed alarms, and timely and efficiently identifies and handles potential threats, thereby improving the safety and stability of the substation.

[0047] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A synchronous monitoring and early warning method for low power consumption of substation ground current, characterized in that, The method includes: By sensing the grounding points of multiple devices in the substation through the substation's grounding network, a ground current dataset is obtained. The acquisition timing sequence is determined by traversing the multiple device grounding points, and the ground current dataset is dynamically time-normalized according to the acquisition timing sequence to obtain a timing synchronization dataset. Time series analysis is performed based on the aforementioned time-series synchronization dataset, and feature extraction is performed based on the analysis results to determine multiple ground current characteristics. The current fluctuation learning is performed by traversing the time-series synchronous dataset according to the multiple ground current features, and a ground current prediction dataset is generated based on the learning results. The grounding network of the substation is dynamically monitored based on the ground current prediction dataset, and multiple feedback response signals are generated to intelligently trigger alarms for the substation. The method involves sensing the grounding points of multiple devices in the substation through the substation's grounding network to obtain ground current data. The substation is covered by the grounding network to obtain multiple data acquisition nodes of the substation. The substation current is captured based on the multiple acquisition nodes, and the grounding points of the multiple devices are determined based on the current capture results. Multiple sensors are activated and the data acquisition frequency is set based on the substation's operating status information; Based on the multiple sensing devices, the grounding points of the multiple devices are traversed and sensed according to the acquisition frequency to obtain multiple ground current data signals, each of which contains a timestamp. The multiple ground current data signals are integrated to determine the ground current dataset; The method includes determining the acquisition timing sequence by traversing the multiple device grounding points, and dynamically time-normalizing the ground current dataset according to the acquisition timing sequence to obtain a timing synchronization dataset. The acquisition timing sequence is determined by traversing the multiple device grounding points and combining the timestamp of each ground current data signal. The acquisition timing sequence is used as a timing reference for node selection to determine the timing baseline; The multiple ground current data signals are amplified to obtain multiple ground current enhancement signals; Based on the timing reference, the multiple ground current enhancement signals are dynamically time-warped to determine the signal time step. The ground current dataset is interpolated and completed according to the signal time step to generate multiple time windows; Consistency verification is performed on the multiple time windows. Based on the verification results, the multiple time windows are associated and integrated to output the time-series synchronization dataset. The method for dynamically time-normalizing the plurality of ground current enhancement signals based on the timing reference to determine the signal time step includes: Based on the time-series reference, the similarity of the multiple ground current enhancement signals is calculated to obtain multiple similarity coefficients; A two-dimensional matrix is ​​constructed based on the time series reference, and Euclidean distance is calculated for the multiple ground current enhancement signals according to the multiple similarity coefficients to generate multiple distance data. The multiple distance data are mapped to the two-dimensional matrix for path optimization to determine the time-aligned path; Based on the timing alignment path, the multiple ground current enhancement signals are time-aligned to generate the signal time step. The method includes learning current fluctuations by traversing the time-series synchronization dataset according to the multiple ground current characteristics, and generating a ground current prediction dataset based on the learning results. Multiple sliding windows are obtained by combining the multiple time windows with the multiple ground current characteristics; By traversing and capturing multiple sliding windows, multiple dynamic current fluctuation data are obtained. The multiple current dynamic fluctuation data are learned using a long short-term memory neural network to generate learning results; Based on the learning results and the time-series synchronization dataset, trend analysis was performed, and a current fluctuation curve was plotted. Based on the current fluctuation curve and the learning results, the training is performed to obtain the training results. Based on the training results and the multiple ground current features, the prediction is performed according to the time series benchmark to generate a ground current prediction data sequence. The geocurrent prediction data sequence is added to the geocurrent prediction dataset.

2. The synchronous monitoring and early warning method for low power consumption of substation ground current as described in claim 1, characterized in that, The grounding network of the substation is dynamically monitored based on the aforementioned ground current prediction dataset, and multiple feedback response signals are generated to intelligently trigger alarms for the substation. The method includes: Multiple operating condition information is retrieved, and the grounding network is started to perform operation simulation analysis based on the multiple operating condition information to obtain multiple ground current operation simulation data. Multiple simulation-prediction data pairs are generated by iterating and matching the multiple ground current operation simulation data with the ground current prediction dataset. The multiple simulation-prediction data pairs are dynamically compared, and multiple feedback response signals are generated based on the comparison results. By iterating through the multiple feedback response signals and combining the multiple simulation-prediction data pairs, current change analysis is performed to generate multiple anomaly labels; The multiple feedback response signals are identified by the multiple abnormal labels, multiple abnormal signals are determined, and intelligent alarms are triggered based on the multiple abnormal signals.

3. The synchronous monitoring and early warning method for low power consumption of substation ground current as described in claim 2, characterized in that, The method involves iterating through the multiple feedback response signals and combining them with the multiple simulation-prediction data pairs to perform current change analysis and generate multiple anomaly labels. Current feedback analysis is performed based on the multiple feedback response signals to determine multiple current change characteristics; Based on the multiple current change characteristics, a current change deviation threshold is set, and the multiple simulation-prediction data pairs are used to calculate the error, generating multiple current change error values. The plurality of current change error values ​​are compared with the current change deviation threshold to determine whether the plurality of current change error values ​​are greater than or equal to the current change deviation threshold. If the multiple current change error values ​​are greater than or equal to the current change deviation threshold, then the simulation-prediction data pairs corresponding to the current change error values ​​that are greater than or equal to the current change deviation threshold are identified, and the multiple anomaly tags are generated.

4. The synchronous monitoring and early warning method for low power consumption of substation ground current as described in claim 2, characterized in that, The method involves identifying multiple feedback response signals based on multiple anomaly tags, determining multiple anomaly signals, and intelligently triggering alarms based on the multiple anomaly signals, including: Using the multiple abnormal labels as indexes, the multiple feedback response signals are traversed and matched to determine the multiple abnormal signals; An impact assessment is performed based on the aforementioned multiple abnormal signals, and multiple impact levels are generated based on the assessment results. The multiple abnormal signals are sorted according to the multiple impact levels to generate signal sorting results; Multiple alarm triggering conditions are set based on the signal screening results, and alarms are intelligently triggered based on the multiple alarm triggering conditions and the multiple abnormal signals.

Images