A method and system for detecting line faults in public transformer substations based on the Internet of Things

Abstract

This invention relates to the field of line fault detection technology, specifically to a method and system for detecting line faults in public transformer substations based on the Internet of Things (IoT). The method includes: acquiring current monitoring signals and triaxial acceleration signals of the public transformer substation during a monitoring period; determining zero-sequence current data and swing amplitude data within the monitoring period based on the data distribution of all current monitoring signals and the data distribution of the triaxial acceleration signals; obtaining coordinated state time point pairs, non-coordinated current point sets, and non-coordinated swing point sets based on the degree of deviation between the zero-sequence current data and swing amplitude data within the monitoring period and the continuous distribution characteristics of time delay in the time sequence; and determining the fault state of the public transformer substation line based on the quantity distribution of coordinated state time point pairs, the distribution of zero-sequence current data in the non-coordinated current point sets, and the quantity distribution of the non-coordinated swing point sets. This invention can obtain more accurate fault types and improve the accuracy of fault detection results.

Description

Technical Field

[0001] This invention relates to the field of line fault detection technology, specifically to a method and system for detecting line faults in public transformer substations based on the Internet of Things. Background Technology

[0002] The public transformer substation is a core link at the end of the power distribution network, and its operational stability directly affects the reliability of power supply at the end. The overhead lines of public transformer substations are the main form of outdoor wiring. A large number of lines are erected in suburban areas, mountainous areas, urban-rural fringe areas, and areas around farmland. These areas often lack effective wind protection and are in a windy and variable wind environment all year round, which becomes a high-frequency scenario for line faults.

[0003] In windy weather, overhead power lines will oscillate periodically under the continuous drive of the wind. This oscillation directly causes two key physical changes: first, the vertical distance between the conductors and the ground will repeatedly increase or decrease with the oscillation; second, the horizontal distance between adjacent conductors will also dynamically change. Overhead power lines inherently possess natural capacitance between themselves and the ground, and between conductors themselves. The magnitude of this capacitance is directly related to the distance between the conductors, and continuous changes in distance will cause the overall capacitance of the line to change dynamically. This dynamic fluctuation in capacitance will generate leakage current in the zero-sequence loop of the line. This current, arising from capacitance changes, is a capacitive leakage current, and its fluctuation frequency will be consistent with the oscillation frequency of the line, while its magnitude will change accordingly with the amplitude of the oscillation.

[0004] In actual power grid operation, this capacitive leakage current caused by line swaying presents a core challenge for fault detection: on the one hand, the amplitude range of this type of capacitive current highly overlaps with the amplitude range of high-resistance grounding leakage current caused by faults such as tree branches crossing the line or minor damage to insulators. Simply put, the values ​​of the two currents are almost the same, and they cannot be effectively distinguished by current magnitude alone; on the other hand, the capacitive current caused by line swaying will appear and disappear depending on the wind force, while the leakage current caused by faults such as tree branches crossing the line or minor damage to insulators will also exhibit intermittent characteristics due to unstable contact. The two currents exhibit highly similar characteristics.

[0005] This makes it difficult for existing fault detection methods, which mainly rely on monitoring whether the amplitude of zero-sequence current exceeds a set threshold, to accurately distinguish the true fault state of abnormal currents in the line. As a result, the accuracy of fault detection results is poor, which creates hidden dangers of misjudgment and omission in subsequent fault judgment. Summary of the Invention

[0006] To address the problem that existing fault detection methods, which rely on monitoring whether the amplitude of zero-sequence current exceeds a set threshold, are unable to accurately distinguish the true fault state of abnormal currents in a line, resulting in poor fault detection accuracy, this invention aims to provide a method and system for detecting line faults in public transformer substations based on the Internet of Things (IoT). The specific technical solution adopted is as follows: In a first aspect, the present invention provides a method for detecting line faults in public transformer substations based on the Internet of Things, comprising: Acquire the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period; Based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals in each monitoring period, the zero-sequence current data and swing amplitude data in each monitoring period are determined respectively. Based on the degree of deviation between the zero-sequence current data and the swing amplitude data in each monitoring period, as well as the continuous distribution characteristics of the time delay in the time sequence, the cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets in each monitoring period are obtained. The fault status of the transformer substation is determined based on the distribution of the number of time points in the coordinated state, the distribution of zero-sequence current data in the non-coordinated current point set, and the distribution of the number of non-coordinated swing points.

[0007] Preferably, the step of obtaining the cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets for each monitoring period based on the degree of deviation between the zero-sequence current data and the swing amplitude data within each monitoring period, as well as the continuous distribution characteristics of the time delay in the time sequence, specifically includes: For any given monitoring period, characteristic constraints are obtained based on the degree of deviation between the zero-sequence current data and the swing amplitude data and the preset deviation threshold, as well as the time delay and the preset delay threshold. Under the condition of satisfying the characteristic constraints, based on the number distribution and time series distribution of matching pairs between zero-sequence current data and swing amplitude data at each time moment, the time point pairs of matching pairs formed by zero-sequence current data and swing amplitude data are filtered to obtain cooperative state time point pairs, non-cooperative current point sets and non-cooperative swing point sets.

[0008] Preferably, the step of obtaining the characteristic constraint conditions based on the degree of deviation between the zero-sequence current data and the swing amplitude data and a preset deviation threshold, and the time delay and a preset delay threshold, specifically includes: The characteristic constraints include amplitude deviation constraints and time delay constraints; The amplitude deviation constraint is specifically defined as follows: the absolute value of the difference between the zero-sequence current data at the first moment and the swing amplitude data at the second moment is less than a preset deviation threshold. The time delay constraint is specifically defined as follows: the time interval between the first moment and the second moment is less than a preset delay threshold. Here, the first moment refers to any moment, and the second moment refers to any moment before the first moment.

[0009] Preferably, under the condition of satisfying the feature constraints, the step of filtering the time point pairs of matching pairs formed by zero-sequence current data and swing amplitude data according to the quantity distribution and time series distribution between the zero-sequence current data and the swing amplitude data at each time point to obtain cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets specifically includes: The zero-sequence current data and the swing amplitude data at each moment constitute each matching pair. The moment corresponding to the zero-sequence current data is the current moment, and the moment corresponding to the swing amplitude data is the swing moment. The current moment and the swing moment constitute the time point pair corresponding to the matching pair. All time point pairs corresponding to the matching pairs that satisfy the feature constraints and have the largest number of consecutive time points are selected as cooperative state time point pairs; The current moments in time points where there is no cooperative state are formed into a set of non-cooperative current points, and the oscillation moments in time points where there is no cooperative state are formed into a set of non-cooperative oscillation points.

[0010] Preferably, the step of determining the fault status of the transformer substation line based on the quantity distribution of coordinated state time point pairs, the zero-sequence current data distribution in the non-coordinated current point set, and the quantity distribution of the non-coordinated swing point set specifically includes: The proportion of the number of cooperative state time point pairs is used as the cooperative characteristic value; the proportion of the sum of squares of all zero-sequence current data in the non-cooperative current point set is used as the non-cooperative energy value; the proportion of the number of all times in the non-cooperative swing point set is used as the loop anomaly characteristic value. Based on the aforementioned non-cooperative energy value, the advanced fault status of the public transformer substation lines is determined; Based on the non-cooperative energy value and the abnormal circuit characteristic value, the intermediate fault state of the public transformer substation line is determined. Based on the non-cooperative energy value, circuit abnormal characteristic value, and cooperative characteristic value, the normal state of the public transformer substation line is determined.

[0011] Preferably, determining the advanced fault status of the transformer substation line based on the non-cooperative energy value specifically includes: When the non-cooperative energy value exceeds the preset energy alarm threshold, the transformer substation is determined to be in an advanced fault state.

[0012] Preferably, determining the intermediate fault state of the transformer substation line based on the non-cooperative energy value and the circuit abnormality characteristic value specifically includes: When the non-cooperative energy value is less than or equal to the energy alarm threshold, and the circuit abnormality characteristic value is greater than the preset missing alarm threshold, the transformer substation line is determined to be in a medium-level fault state.

[0013] Preferably, determining the normal state of the transformer substation lines based on the non-cooperative energy value, circuit anomaly characteristic value, and cooperative characteristic value specifically includes: When the non-cooperative energy value is less than or equal to the energy alarm threshold, the circuit abnormality characteristic value is less than or equal to the missing alarm threshold, and the cooperative characteristic value is greater than the preset wind deflection cooperative threshold, the transformer substation line is determined to be in a normal state.

[0014] Preferably, the step of determining the zero-sequence current data and oscillation amplitude data for each monitoring period based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period specifically includes: For any given monitoring period, the L2 norm of the triaxial acceleration within the monitoring period is standardized to obtain the swing amplitude data for that period. The root mean square of the current within the sliding window is calculated based on the high-pass filtered data of the current monitoring signal during the monitoring period. The zero-sequence current data for the monitoring period is obtained by standardizing the result of low-pass filtering of the root mean square current.

[0015] Secondly, the present invention provides an Internet of Things (IoT)-based public transformer substation line fault detection system. This system implements the steps of an IoT-based public transformer substation line fault detection method. The IoT-based public transformer substation line fault detection system includes: The data acquisition module is used to acquire the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period; The data preprocessing module is used to determine the zero-sequence current data and swing amplitude data for each monitoring period based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period. The collaborative feature analysis module is used to obtain the collaborative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets for each monitoring period based on the degree of deviation between the zero-sequence current data and swing amplitude data in each monitoring period and the continuous distribution characteristics of the time delay in the time sequence. The fault status detection module is used to determine the fault status of the transformer substation lines based on the quantity distribution of coordinated state time point pairs, the zero-sequence current data distribution in the non-coordinated current point set, and the quantity distribution of the non-coordinated swing point set.

[0016] The embodiments of the present invention have at least the following beneficial effects: This invention first acquires current monitoring signals (reflecting the electrical state of the line) and triaxial acceleration signals (reflecting the mechanical oscillation state of the line) for each monitoring period to ensure the acquisition of raw data on both the electrical and physical dimensions of line operation. This provides a complete and undistorted data foundation for subsequent feature extraction and fault analysis. Then, for the current monitoring signals, zero-sequence current data is extracted through data processing to accurately capture changes in line leakage current. For the triaxial acceleration signals, oscillation amplitude data is calculated through data processing to intuitively reflect the intensity of line mechanical oscillation. Furthermore, based on the amplitude deviation and time delay of the zero-sequence current data and oscillation amplitude data, time point pairs representing normal wind deflection are selected, reflecting the natural correlation between oscillation and current. Simultaneously, non-cooperative current point sets are separated, reflecting abnormal currents without oscillation, and non-cooperative oscillation point sets are separated, reflecting abnormal oscillations without corresponding current responses. This step resolves the confusion between wind deflection interference and fault signals, providing clear abnormal characteristic basis for subsequent fault determination. Finally, by analyzing the distribution of the number of time points in the cooperative state, the distribution of zero-sequence current data in the non-cooperative current point set, and the distribution of the number of non-cooperative oscillation point sets, different fault states are accurately identified, including high-risk faults such as tree obstructions or foreign object connections, and medium-risk hidden dangers such as abnormal circuit impedance. This step avoids false alarms and missed alarms caused by simply relying on amplitude thresholds, obtains more accurate fault types, improves the accuracy of fault detection results, and provides direct decision support for precise operation and maintenance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the steps of a public transformer substation line fault detection method based on the Internet of Things provided by the present invention; Figure 2 This is a structural block diagram of a public transformer substation line fault detection system based on the Internet of Things provided by the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a public transformer substation line fault detection method and system based on the Internet of Things proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] 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 invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details the specific scheme of the Internet of Things-based public transformer substation line fault detection method and system provided by the present invention.

[0022] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting line faults in a public transformer substation based on the Internet of Things, according to an embodiment of the present invention. The method includes the following steps: Step S100: Obtain the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period.

[0023] In this embodiment, a monitoring period refers to a time range including the moment when vibration exceeds the limit and the preset time length thereafter. Specifically, when the triaxial accelerometer detects that the composite acceleration amplitude is greater than or equal to a preset wind threshold, the current moment is determined to be the moment when vibration exceeds the limit. The composite acceleration amplitude refers to the magnitude of the vector sum of the three orthogonal components output by the triaxial accelerometer in Euclidean space. The moment when vibration exceeds the limit indicates that the line has transitioned from a relatively static state to a state of significant oscillation.

[0024] As a concrete example, regarding the moment when vibration exceeds the limit... The corresponding monitoring period can be represented as This time frame includes the line oscillation process and the possible accompanying fault discharge process, and is the main object of fault characteristic analysis.

[0025] In other embodiments, the moment when the vibration exceeds the limit can also be recorded. The corresponding background data and the corresponding time range are: The data within this time frame includes the environmental noise and micro-motion status of the line before the large swing occurs, which can be used as background data for data normalization processing.

[0026] Furthermore, during the monitoring period, a triaxial accelerometer is used to collect triaxial acceleration signals, which include three orthogonal components. The sampling frequency can be set to 100Hz, which can be adjusted by the implementer according to the specific implementation scenario. A zero-sequence current transformer (ZCT) is used to collect current monitoring signals. Its data acquisition frequency is usually higher, such as 4000Hz, to capture high-frequency discharge characteristics. This frequency can also be adjusted by the implementer according to the specific implementation scenario.

[0027] Step S200: Based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period, determine the zero-sequence current data and swing amplitude data within each monitoring period.

[0028] Considering that the sampling frequency of current monitoring signals is usually high and includes 50Hz power frequency components and various high-frequency noises, while the mechanical oscillation frequency of the line is usually low, the corresponding triaxial acceleration signal has a low sampling frequency. If the current monitoring signal and the triaxial acceleration signal are directly used for subsequent feature matching analysis, the huge difference in their frequency bands will cause the feature matching process to fail. Therefore, the main purpose of this step is to extract key low-frequency features and achieve time dimension alignment between the current monitoring signal and the triaxial acceleration signal.

[0029] The first step is to standardize and dimensionlessly process the triaxial acceleration signals.

[0030] Specifically, in this embodiment, all feature analysis operations starting from this step are described for any given monitoring period. The L2 norm of the triaxial accelerations within the monitoring period is standardized to obtain the oscillation amplitude data for that period. It should be noted that the triaxial accelerometer data collected at each moment includes three orthogonal components. The L2 norm of the three orthogonal components is calculated, and Z-score standardization is performed to obtain the swing amplitude data at each moment within the monitoring period. In other embodiments, the mean and variance data used for standardization can be obtained statistically from the background data. The standardization method is a well-known technique and will not be described in detail here. To avoid calculation errors, a preset very small positive number (such as 0.0001) is added to the denominator of the variance term in the Z-score standardization to prevent division by zero errors triggered when the signal is extremely stable or there is no current response.

[0031] The second step is to standardize and dimensionlessly process the current monitoring signal.

[0032] Specifically, based on the current monitoring signal during the monitoring period, the root mean square of the current within the sliding window is calculated after high-pass filtering; the result of low-pass filtering of the root mean square of the current is normalized to obtain the zero-sequence current data during the monitoring period.

[0033] As a specific example, in this embodiment, the length of the sliding window is set to the ratio between the sampling frequency of the current monitoring signal and the sampling frequency of the triaxial acceleration signal. The root mean square (RMS) of the current in each sliding window is calculated after high-pass filtering. The RMS of the current in all sliding windows within the monitoring period is then low-pass filtered and followed by Z-score normalization to obtain the preprocessed zero-sequence current data.

[0034] It should be understood that a sliding window corresponds to one zero-sequence current data point. In this case, the lengths of the swing amplitude data and the zero-sequence current data are the same, and the time intervals between each point are identical, achieving timing alignment. It should be noted that high-pass and low-pass filtering operations are well-known techniques and will not be elaborated upon here.

[0035] Step S300: Based on the degree of deviation between the zero-sequence current data and the swing amplitude data in each monitoring period and the continuous distribution characteristics of the time delay in the time sequence, obtain the cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets in each monitoring period.

[0036] The main purpose of this step is to quantify the cooperative and non-cooperative relationship between zero-sequence current and line sway. Through dual constraints and time-series screening, the cooperative component driven by normal wind deflection and the non-cooperative component related to faults are accurately separated, providing clear characteristic basis for subsequent fault classification judgment.

[0037] In windy environments, zero-sequence current and sway data exhibit amplitude overlap, time lag, and discrete noise interference. Relying solely on a single threshold cannot distinguish between "wind-driven current" and "fault abnormal current," nor can it identify loop anomalies such as "swaying without current." Therefore, this step achieves feature separation through a three-level logical progression. First, based on the degree of deviation between zero-sequence current and sway amplitude (matching the rationality of physical amplitude) and time delay (matching physical response timing), dual feature constraints are set to eliminate invalid matches that clearly do not meet physical laws. Then, within the constraints, combining the quantity distribution of matching pairs (ensuring statistical significance) and time sequence distribution (ensuring physical continuity), continuous and consistent coordinated state time point pairs (representing normal wind deflection) are selected. Finally, unmatched current data is stripped away to form a non-coordinated current point set, representing externally injected faults such as tree obstacles, and unmatched sway data is selected to form a non-coordinated sway point set, representing abnormal loop impedance. This achieves the core objective of accurately extracting fault features from mixed signals.

[0038] In response, firstly, for any given monitoring period, characteristic constraints are derived based on the degree of deviation between the zero-sequence current data and the swing amplitude data and a preset deviation threshold, as well as the time delay and a preset delay threshold.

[0039] Specifically, the feature constraints include amplitude deviation constraints and time delay constraints; the amplitude deviation constraint is specifically: the absolute value of the difference between the zero-sequence current data at the first moment and the swing amplitude data at the second moment is less than a preset deviation threshold; the time delay constraint is specifically: the time interval between the first moment and the second moment is less than a preset delay threshold; wherein, the first moment refers to any moment, and the second moment refers to any moment before the first moment.

[0040] As a specific example, in this embodiment, the deviation threshold can be set to 1.5. This value is an empirical value obtained through a large number of experiments, and the implementer can set it according to the specific implementation scenario. The amplitude deviation constraint is used to determine whether the amplitude of the zero-sequence current data at the first moment and the swing amplitude data at the second moment are consistent within the allowable tolerance range. When the amplitude deviation constraint is met, it indicates that the response intensity of the zero-sequence current of the line matches the intensity of the line swing, which conforms to the physical law that the larger the swing amplitude, the greater the change in conductor spacing or distance to ground, and the greater the change in line distributed capacitance. The purpose of using the threshold is to allow for small deviations between the swing and current amplitudes, such as sensor noise or minor environmental interference, but the deviation cannot exceed the reasonable allowable range.

[0041] In this embodiment, the delay threshold can be 200 milliseconds. This value is an empirical value obtained through extensive experimental statistics, and the implementer can set it according to the specific implementation scenario. The time delay constraint is used to determine whether the misalignment between the two on the time axis is within the physically permissible delay range. When the time delay constraint is met, it indicates that the time of the zero-sequence current response and the time of the line swing are synchronized, conforming to the physical timing characteristics of the swing occurring first, the capacitance change occurring later, and the current occurring later or almost simultaneously. The purpose of using the threshold is to allow a slight time lag in the current response, but the lag cannot exceed the reasonable allowable response range.

[0042] It should be noted that the aforementioned amplitude deviation constraints and time delay constraints are statistical empirical boundaries extracted from historical operating data. The numerical comparison after Z-Score processing aims to establish a characteristic correlation mapping between wind deflection and current at the level of data distribution deviation, allowing for reasonable physical microwave disturbances, rather than requiring an absolute physical equivalence conversion between the two.

[0043] Secondly, under the condition of satisfying the characteristic constraints, based on the number distribution and time series distribution of matching pairs between the zero-sequence current data and the swing amplitude data at each time moment, the time point pairs of matching pairs formed by the zero-sequence current data and the swing amplitude data are filtered to obtain the cooperative state time point pairs, the non-cooperative current point set, and the non-cooperative swing point set.

[0044] Specifically, all time point pairs that satisfy the feature constraints and have the largest number of consecutive matching pairs in time sequence are selected as cooperative state time point pairs; the current moments in time point pairs that do not exist in cooperative state form a non-cooperative current point set, and the oscillation moments in time point pairs that do not exist in cooperative state form a non-cooperative oscillation point set.

[0045] In this context, the zero-sequence current data and the swing amplitude data at each moment constitute each matching pair, with the moment corresponding to the zero-sequence current data as the current moment and the moment corresponding to the swing amplitude data as the swing moment; the current moment and the swing moment constitute the time point pair corresponding to the matching pair.

[0046] It should be noted that, due to the nonlinear characteristics of wind-driven line oscillations, the oscillation period may vary slightly with wind speed, resulting in a non-fixed time lag or lead in the current response relative to the oscillation waveform. Traditional linear correlation analysis (such as Pearson coefficients) requires strict waveform alignment and cannot handle this local time shift. Therefore, this step employs the Longest Common Subsequence (LCSS) algorithm to determine the cooperative relationship between oscillation and current point by point.

[0047] To address this, this embodiment utilizes the core logic of the Longest Common Subsequence (LCSS) algorithm. Through iterative construction of a two-dimensional matrix and reverse backtracking, it accurately selects the matching pairs that satisfy the feature constraints and have the largest number of consecutive temporal sequences, ultimately extracting the cooperative state time point pairs. The specific process is as follows: As a concrete example, to achieve accurate determination of the cooperative relationship between oscillation and nonlinear time shift of current, this embodiment first constructs a two-dimensional dynamic programming matrix for quantifying the temporal cooperative length, which serves as the core computational carrier of the LCSS algorithm. The size of the matrix is... N represents the total number of zero-sequence current data points, which is also the total number of oscillation amplitude data points. The row index of the matrix represents the time index corresponding to the oscillation amplitude data, and the column index represents the time index corresponding to the zero-sequence current data. It should be noted that the row index of the matrix can also refer to the index of the oscillation time, and the column index can also refer to the index of the current time. Each position in the matrix corresponds to a matching pair, and the current time and oscillation time corresponding to each position correspond to a time point pair.

[0048] Elements in the matrix This is used to record the maximum continuous cooperative length that can be formed by matching pairs that satisfy the characteristic constraints between the first i moments of the oscillation amplitude data in chronological order and the first j moments of the zero-sequence current data. All elements in the first row (corresponding to the moment without oscillation) and the first column (corresponding to the moment without current) of the two-dimensional matrix are uniformly assigned the value 0, thereby establishing a physical benchmark of no response without input, providing a zero reference dimension for subsequent iterative calculations.

[0049] After completing matrix initialization, this embodiment starts from... , The matrix elements are iteratively updated point by point. Throughout the iteration process, the characteristic constraints of amplitude deviation and time delay constraints are used as the basis for judgment, regarding the matrix position... The matching pair formed by the i-th oscillation moment and the j-th current moment, if the matching pair satisfies the characteristic constraint condition, indicates that it conforms to the physical law of current response induced by wind-driven line oscillation. At this time, Assign the value to the top-left adjacent element The length of the current consecutive collaboration is incremented by 1 based on the previous optimal result, indicating that the matching pair is an effective extension of the previous consecutive collaboration sequence; if the matching pair does not satisfy the feature constraint, it means that it is a non-homologous mismatch pair, and in this case, it will be... Assigned to the upper element With the element on the left The maximum value in the sequence is used to inherit the maximum continuous cooperative length found in the previous iteration, thus avoiding the interruption of the determination of the global optimal cooperative sequence due to a single mismatch point.

[0050] After constructing the entire matrix through the above point-by-point iterative operations, the final element of the matrix... The value represents the longest continuous collaboration length that can be formed by matching pairs that satisfy the feature constraints throughout the entire monitoring period. Based on this, starting from the matrix endpoint... Begin the reverse backtracking derivation, based on the source of the matrix elements' values, i.e., determine... It is by Updated, or by The update is derived by tracing back point by point to form the temporal path of the longest continuous cooperative length, only including the matrix positions "updated from L[i-1,j-1]+1" during the backtracking process. The matching pair formed by the i-th oscillation moment and the j-th current moment is included in the screening range. The time point pairs corresponding to the finally extracted matching pairs are the cooperative state time point pairs that satisfy the feature constraints and have the largest number of consecutive time sequences.

[0051] The cooperative state time point pair represents the homogeneous response that is truly driven by continuous wind (the wind swing is continuous and periodic, and the corresponding current response must also be continuous). The selection of the most continuous number is to maximize the preservation of normal cooperative components and minimize noise interference.

[0052] Based on all coordinated state time point pairs, if a certain current moment does not appear in all coordinated state time point pairs, it means that there is no corresponding swing driving source for that current moment. Therefore, the current moment is classified into the non-coordinated current point set, indicating an abnormal current without swing driving. This means that the abnormality of the zero-sequence current data in the point set is not caused by wind swing, so a matching swing moment cannot be found. In practice, it may be leakage current caused by faults such as tree obstacles or insulator damage. It usually corresponds to the discharge at the moment of tree obstacle contact or the pulse generated by foreign object connection. Its characteristic is "non-homogeneous morphology".

[0053] Based on all coordinated state time points, if a certain swing moment does not appear in all coordinated state time points, it means that the swing did not produce a corresponding current response. Therefore, the swing moment is classified into the non-coordinated swing point set, indicating that the line actually swings (the capacitance should change), but no corresponding current response is produced due to the circuit abnormality. Therefore, it is impossible to find a matching current moment. In practice, it may be due to abnormal circuit impedance, loose hardware, or other abnormal situations. Usually, it corresponds to unstable contact caused by loose hardware (i.e., the circuit is momentarily open when swinging to a specific angle, blocking the current) or the sensor itself falling off.

[0054] Thus, the coordinated state time point pair characterizes the time interval corresponding to the zero-sequence capacitive leakage current generated by the dynamic change of distributed capacitance caused by the line swing driven by normal wind force during the monitoring period. The current response in this interval is the physical result of the normal operation of the line, without fault-related abnormal signal interference, and is the core characteristic basis for subsequent judgment of fault-free wind swing.

[0055] Non-cooperative current point set characterizes the time sequence point set corresponding to abnormal zero-sequence currents that are not driven by line wind sway during the monitoring period. These currents usually originate from high-resistance grounding faults such as tree obstruction, minor insulator damage, or non-wind sway factors such as sensor noise and external electromagnetic interference. They are the core abnormal features for determining current-side faults such as tree obstruction.

[0056] The non-cooperative swing point set represents the time series point set in which, although the line experiences mechanical swing (which should theoretically trigger capacitance changes and current responses) during the monitoring period, no corresponding zero-sequence current response is generated due to problems such as abnormal circuit impedance, loose fittings, and abnormal line grounding status. It is the core feature for determining abnormal electrical circuits of the line and reflects the link failure between mechanical swing and electrical response.

[0057] Step S400: Determine the fault status of the transformer substation line based on the quantity distribution of the coordinated state time point pairs, the zero-sequence current data distribution in the non-coordinated current point set, and the quantity distribution of the non-coordinated swing point set.

[0058] The main purpose of this step is to transform the discrete point set after feature separation into quantitative indicators, accurately determine the operating status of the transformer substation lines through hierarchical decision logic, realize differentiated identification and alarm for faults of different severity levels, and solve the decision conflict problem when multiple features occur concurrently.

[0059] In windy environments, power lines may simultaneously experience wind-induced deviation interference, external fault injection, and internal circuit anomalies. Different faults have different hazard priorities (external faults such as tree obstructions > circuit impedance anomalies > normal wind deviation). Simply summarizing these characteristics is insufficient for making clear maintenance decisions. Therefore, this step first quantifies the coordinated state time point pairs, non-coordinated current point sets, and non-coordinated swing point sets into directly identifiable indicators: coordinated characteristic values ​​(representing the proportion of normal wind deviation), non-coordinated energy values ​​(representing the intensity of external fault injection), and circuit anomaly characteristic values ​​(representing the degree of internal connection hazards). Then, following a hazard-priority grading rule, high-risk advanced faults (tree obstructions or foreign object connections) are prioritized based on non-coordinated energy values. Medium-risk intermediate faults (circuit impedance anomalies) are then determined based on both non-coordinated energy values ​​and circuit anomaly characteristic values. Finally, the normal wind deviation status is comprehensively determined by combining the three indicators. The final output provides a clear line operating status and maintenance priority, ensuring that high-risk faults are not overlooked and avoiding false alarms of normal wind deviation, providing a clear basis for accurate maintenance.

[0060] Specifically, the proportion of the number of cooperative state time point pairs is used as the cooperative characteristic value; the proportion of the sum of squares of all zero-sequence current data in the non-cooperative current point set is used as the non-cooperative energy value; and the proportion of the number of all times in the non-cooperative swing point set is used as the loop anomaly characteristic value.

[0061] As a concrete example, the ratio of the total number of all coordinated state time point pairs to the total number of all times within the monitoring period is used to obtain the coordinated characteristic value. The closer this value is to 1, the higher the coordination between the current response and the oscillation, and the more the line condition tends to be in normal wind deflection.

[0062] The sum of the squares of all zero-sequence current data in the non-cooperative current point set is recorded as the first sum, and the sum of the squares of all zero-sequence current data within the monitoring period is recorded as the second sum. The ratio of the first sum to the second sum is the non-cooperative energy value, which reflects the proportion of non-cooperative current energy. The larger this value, the stronger the non-cooperative current injection in the line, usually corresponding to the discharge energy generated by tree collisions or foreign object contact. It should be noted that, to avoid the ratio calculation failure, a preset very small positive number (such as 0.0001) is added to prevent the error of division by zero triggered when the signal is extremely stable or there is no current response.

[0063] The ratio between the total number of times in the non-cooperative swing point set and the total number of times corresponding to the swing amplitude data is used to obtain the circuit anomaly characteristic value. The larger this value is, the more frequently the line experiences "movement without power" during the swing process, suggesting that there may be an anomaly in the zero-sequence circuit impedance.

[0064] Furthermore, in this embodiment, three preset judgment thresholds are defined as an energy alarm threshold, a missing alarm threshold, and a wind deflection coordination threshold. The energy alarm threshold ranges from 0.2 to 0.4. For example, a value of 0.3 indicates that the non-coordinated energy proportion is 30%. The missing alarm threshold can range from 0.15 to 0.25 (meaning that 15% to 25% of vibration moments do not match a current response), preferably 0.2, which can be set by the implementer according to the specific implementation scenario. The wind deflection coordination threshold can range from 0.6 to 0.8 (meaning that 60% to 80% of current fluctuations are driven by wind deflection), preferably 0.7, which can be set by the implementer according to the specific implementation scenario.

[0065] The first step is to determine the advanced fault status of the transformer substation lines based on the non-cooperative energy values.

[0066] Specifically, when the non-cooperative energy value exceeds the preset energy alarm threshold, the transformer substation is determined to be in a high-level fault state. At this point, regardless of the performance of the cooperative characteristic value, any high-energy non-homogeneous injection is considered a high-risk fault. The current state is determined to be a tree obstruction or a potential foreign object connection hazard, and a level-one alarm work order is generated.

[0067] The second step is to determine the intermediate fault status of the transformer substation lines based on the non-cooperative energy value and the abnormal circuit characteristic value. Specifically, when the non-cooperative energy value is less than or equal to the energy alarm threshold and the circuit abnormality characteristic value is greater than the preset missing alarm threshold, the transformer substation line is determined to be in a medium-level fault state.

[0068] At this point, after ruling out external discharge, if a significant amount of oscillation fails to elicit an electrical response, it indicates an intermittent interruption or impedance surge in the electrical connection. The current state is determined to be a potential loop impedance anomaly, with possible causes including loose fittings or poor wire contact, generating a level-two warning work order.

[0069] The third step is to determine the normal state of the transformer substation lines based on the non-cooperative energy value, circuit abnormality characteristic value, and cooperative characteristic value.

[0070] Specifically, when the non-coordinated energy value is less than or equal to the energy alarm threshold, the circuit abnormal characteristic value is less than or equal to the missing alarm threshold, and the coordinated characteristic value is greater than the preset wind deflection coordinated threshold, the transformer substation line is determined to be in a normal state. At this time, current fluctuations are mainly driven by mechanical oscillation and have no significant abnormal components. This is determined to be normal wind deflection galloping, the event is recorded, but no alarm is triggered.

[0071] In other embodiments, if none of the above threshold conditions are met, that is, when the non-cooperative energy value is less than or equal to the energy alarm threshold, the loop abnormal characteristic value is less than or equal to the missing alarm threshold, and the cooperative characteristic value is less than or equal to the preset wind deflection cooperative threshold, the monitoring period can be marked as a data abnormal event to remind relevant maintenance personnel to check the sensor status.

[0072] In other embodiments, the fault location of a line can be determined based on the fault status detection results at multiple monitoring points on the same line. Specifically, when multiple monitoring points on the same line simultaneously exhibit fault states, if the fault state is determined to be high-level (i.e., a tree obstruction hazard), the monitoring point corresponding to the maximum non-cooperative energy value is identified as the fault location. If the fault state is determined to be medium-level (i.e., a loop anomaly), the monitoring point corresponding to the maximum loop anomaly characteristic value is identified as the fault location.

[0073] It should be noted that a monitoring point refers to a data monitoring location on a line. This embodiment does not involve the characteristics of multiple monitoring points, that is, it only performs feature analysis on the current and oscillation of a single monitoring point on the line.

[0074] like Figure 2 As shown, this embodiment of the invention also provides an Internet of Things (IoT)-based public transformer substation line fault detection system. This system is used to implement the steps of an IoT-based public transformer substation line fault detection method. The IoT-based public transformer substation line fault detection system includes: The data acquisition module is used to acquire the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period; The data preprocessing module is used to determine the zero-sequence current data and swing amplitude data for each monitoring period based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period. The collaborative feature analysis module is used to obtain the collaborative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets for each monitoring period based on the degree of deviation between the zero-sequence current data and swing amplitude data in each monitoring period and the continuous distribution characteristics of the time delay in the time sequence. The fault status detection module is used to determine the fault status of the transformer substation lines based on the quantity distribution of coordinated state time point pairs, the zero-sequence current data distribution in the non-coordinated current point set, and the quantity distribution of the non-coordinated swing point set.

[0075] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting line faults in public transformer substations based on the Internet of Things, characterized in that, The method includes the following steps: Acquire the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period; Based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals in each monitoring period, the zero-sequence current data and swing amplitude data in each monitoring period are determined respectively. Based on the degree of deviation between the zero-sequence current data and the swing amplitude data in each monitoring period, as well as the continuous distribution characteristics of the time delay in the time sequence, the cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets in each monitoring period are obtained. Based on the distribution of the number of time points in the coordinated state, the distribution of zero-sequence current data in the non-coordinated current point set, and the distribution of the number of non-coordinated swing point sets, the fault status of the transformer substation area is determined. Based on the degree of deviation between the zero-sequence current data and the swing amplitude data within each monitoring period, and the continuous distribution characteristics of the time delay in the time sequence, the coordinated state time point pairs, non-coordinated current point sets, and non-coordinated swing point sets within each monitoring period are obtained, specifically including: For any given monitoring period, feature constraints are obtained based on the degree of deviation between the zero-sequence current data and the swing amplitude data and a preset deviation threshold, as well as the time delay and a preset delay threshold. Specifically, the feature constraints include amplitude deviation constraints and time delay constraints. The amplitude deviation constraint is specifically defined as follows: the absolute value of the difference between the zero-sequence current data at the first moment and the swing amplitude data at the second moment is less than a preset deviation threshold. The time delay constraint is specifically defined as follows: the time interval between the first moment and the second moment is less than a preset delay threshold. Here, the first moment refers to any given moment, and the second moment refers to any moment prior to the first moment. Under the condition of satisfying the feature constraints, based on the number distribution and time series distribution of matching pairs between the zero-sequence current data and the swing amplitude data at each time moment, the time point pairs of matching pairs formed by the zero-sequence current data and the swing amplitude data are filtered to obtain cooperative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets. Specifically, each time point pair is formed by the zero-sequence current data and the swing amplitude data at each time moment, with the time moment corresponding to the zero-sequence current data as the current time moment and the time moment corresponding to the swing amplitude data as the swing time moment; the current time moment and the swing time moment constitute the time point pair corresponding to the matching pair; all time point pairs corresponding to all matching pairs that satisfy the feature constraints and have the largest number of consecutive time points are selected as cooperative state time point pairs; the current time moments that do not exist in cooperative state time point pairs constitute the non-cooperative current point set, and the swing time moments that do not exist in cooperative state time point pairs constitute the non-cooperative swing point set. The method of determining the fault status of the transformer substation lines based on the distribution of the number of coordinated state time point pairs, the distribution of zero-sequence current data in the non-coordinated current point set, and the distribution of the number of non-coordinated swing point sets specifically includes: The ratio between the total number of all coordinated state time point pairs and the total number of all times within the monitoring period is used to obtain the coordinated characteristic value; the sum of the squares of all zero-sequence current data in the non-coordinated current point set is recorded as the first sum, and the sum of the squares of all zero-sequence current data within the monitoring period is recorded as the second sum. The ratio of the first sum to the second sum is the non-coordinated energy value; the ratio between the total number of all times in the non-coordinated swing point set and the total number of times corresponding to the swing amplitude data is used to obtain the loop anomaly characteristic value. Based on the aforementioned non-cooperative energy value, the advanced fault status of the public transformer substation lines is determined; Based on the non-cooperative energy value and the abnormal circuit characteristic value, the intermediate fault state of the public transformer substation line is determined. Based on the non-cooperative energy value, circuit abnormal characteristic value, and cooperative characteristic value, the normal state of the public transformer substation lines is determined. The determination of the advanced fault status of the transformer substation lines based on the non-cooperative energy value specifically includes: When the non-cooperative energy value is greater than the preset energy alarm threshold, the public transformer area line is determined to be in a high-level fault state, and the current state is determined to be a tree obstruction or foreign object connection hazard. The determination of the intermediate fault status of the transformer substation lines based on the non-cooperative energy value and the circuit abnormality characteristic value specifically includes: When the non-cooperative energy value is less than or equal to the energy alarm threshold, and the circuit abnormal characteristic value is greater than the preset missing alarm threshold, the transformer substation line is determined to be in a medium-level fault state, and the current state is determined to be a potential circuit impedance abnormality.

2. The method for detecting line faults in public transformer substations based on the Internet of Things according to claim 1, characterized in that, The determination of the normal state of the transformer substation lines based on the non-cooperative energy value, circuit anomaly characteristic value, and cooperative characteristic value specifically includes: When the non-cooperative energy value is less than or equal to the energy alarm threshold, the circuit abnormality characteristic value is less than or equal to the missing alarm threshold, and the cooperative characteristic value is greater than the preset wind deflection cooperative threshold, the transformer substation line is determined to be in a normal state.

3. The method for detecting line faults in public transformer substations based on the Internet of Things according to claim 1, characterized in that, The method, based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period, determines the zero-sequence current data and oscillation amplitude data within each monitoring period, specifically including: For any given monitoring period, the L2 norm of the triaxial acceleration within the monitoring period is standardized to obtain the swing amplitude data for that period. The root mean square of the current within the sliding window is calculated based on the high-pass filtered data of the current monitoring signal during the monitoring period. The zero-sequence current data for the monitoring period is obtained by standardizing the result of low-pass filtering of the root mean square current.

4. A public transformer substation line fault detection system based on the Internet of Things, characterized in that, This system is used to implement the steps of the Internet of Things (IoT)-based public transformer substation line fault detection method as described in any one of claims 1-3, wherein the IoT-based public transformer substation line fault detection system comprises: The data acquisition module is used to acquire the current monitoring signal and triaxial acceleration signal of the public transformer area during each monitoring period; The data preprocessing module is used to determine the zero-sequence current data and swing amplitude data for each monitoring period based on the data distribution of all current monitoring signals and the data distribution of triaxial acceleration signals within each monitoring period. The collaborative feature analysis module is used to obtain the collaborative state time point pairs, non-cooperative current point sets, and non-cooperative swing point sets for each monitoring period based on the degree of deviation between the zero-sequence current data and swing amplitude data in each monitoring period and the continuous distribution characteristics of the time delay in the time sequence. The fault status detection module is used to determine the fault status of the transformer substation lines based on the quantity distribution of coordinated state time point pairs, the zero-sequence current data distribution in the non-coordinated current point set, and the quantity distribution of the non-coordinated swing point set.

Images