High-voltage line fault identification method based on phasor analysis

Abstract

The invention relates to the technical field of line fault analysis, in particular to a high-voltage line fault identification method based on phasor analysis, which comprises the following steps of: performing multi-time window nesting decomposition on voltage phasors and current phasors acquired in the operation process of a high-voltage line, extracting a phasor dynamic response sequence, and calculating the phasor dynamic response sequence; identifying a main disturbance mechanism characteristic triggered by a fault, and constructing a main disturbance mechanism type label; a driving response dominant phasor disturbance factor is identified, a phasor factor mapping graph is constructed, factor stripping operation is executed, and a dominant purification phasor set is generated; and performing phase evolution analysis on the dominant purification phasor set, and determining a fault initial phase transition boundary and a cross-phase offset trend. According to the method, the stability and the physical consistency of identification are ensured, and the finally output fault type and phase structured result not only has clear logic traceability, but also highly meets the requirements of dispatching automation and quick response of a protection system.

Description

Technical Field

[0001] This invention relates to the field of line fault analysis technology, and in particular to a high-voltage line fault identification method based on phasor analysis. Background Technology

[0002] In modern power systems, high-voltage transmission lines are important channels for long-distance power transmission, and their operational stability is directly related to the safety of the entire power grid. However, high-voltage lines are susceptible to interference from factors such as lightning strikes, tree obstructions, equipment aging, and insulation breakdown during long-term operation, resulting in frequent grounding, short circuit, and cross-phase faults. If the faulty phase and fault type cannot be identified in a timely and accurate manner, it will not only delay fault isolation and restoration operations, but may also trigger cascading trips.

[0003] Traditional fault identification methods mostly rely on the setpoint action of relay protection devices, identification of sudden changes in power frequency component amplitude, or harmonic distortion judgment. Many traditional algorithms depend on the judgment of amplitude exceeding limits after steady state, making it difficult to respond quickly in the early stages of a fault, especially in transient and gradual fault scenarios, where the accuracy and timeliness of identification are significantly insufficient. Although fault disturbance processes are often accompanied by harmonic amplification, interharmonic excitation, or high-frequency fluctuations, existing methods lack in-depth analysis of multi-frequency band phasor responses and cannot effectively extract the correlation between frequency domain features and fault mechanisms. Multiphase coupling propagation (such as a fault in phase A causing a linkage in phase B) is very common in practice, but most methods only focus on single-phase abnormal signals, lacking the characterization and judgment of the interphase linkage evolution trend, and are prone to misjudging as single-point faults or missing associated faults. Summary of the Invention

[0004] This invention provides a high-voltage line fault identification method based on phasor analysis. It integrates multi-timescale dynamic response, accurate extraction of dominant disturbance factors, identification of phase offset paths, and determination of fault phase and type to adapt to the increasingly complex power system operating environment and higher fault response requirements.

[0005] A high-voltage line fault identification method based on phasor analysis includes the following steps:

[0006] S1, Disturbance Mechanism Identification and Feature Classification: The voltage and current phasors collected during the operation of high-voltage lines are decomposed using a multi-time-window nested decomposition method to extract the dynamic response sequence of the phasors. Based on the characteristics of the disturbance frequency spectrum and the consistency of the response direction, the dynamic response sequence of the phasors is classified into mechanisms to identify the main disturbance mechanism features triggered by the fault and to construct a main disturbance mechanism type label. The main disturbance mechanism features include power frequency transients, harmonic amplification response and interphase coupling migration behavior, which are used for subsequent extraction of dominant factors.

[0007] S2, Dominant Factor Extraction and Phasor Purification: The main perturbation mechanism features are input into a multi-channel discrimination network. Under the guidance of the main perturbation mechanism classification label, the phasor perturbation factors that drive the response are identified. A phasor factor mapping map is constructed and a factor stripping operation is performed to generate a dominant purified phasor set. The dominant purified phasor set is used for subsequent phase transition identification.

[0008] S3, Phase transition identification and fault output: Perform phase evolution analysis on the dominant purification phasor set, identify the phase angle transition point of the phasor and the abrupt change region of the phase difference, determine the fault initiation phase transition boundary and cross-phase offset trend, and output the fault to which the phase and type identification result belongs based on the trend evolution direction and the degree of synchronization.

[0009] Optionally, the extraction of the phasor dynamic response sequence specifically includes:

[0010] A progressive time window structure is used to perform multi-scale analysis on voltage and current phasor data collected by the synchronous phasor measurement device. The progressive time window structure includes a millisecond-level short window for capturing transient processes, a cycle-level medium window for identifying steady-state shifts, and a hundred-millisecond-level long window for observing trend evolution. By parallel calculation of the phasor amplitude and phase angle change trajectories within each time window, a phasor dynamic response sequence that can characterize the dynamic response of the entire fault process is generated.

[0011] Optionally, in S1:

[0012] The phasor dynamic response sequence is subjected to a fast Fourier transform to extract its perturbation frequency spectrum characteristics. By calculating the energy integral and fundamental energy ratio of a specific frequency band, the harmonic amplification response characteristics are identified. The specific frequency band includes the second harmonic, the third harmonic, and interharmonics.

[0013] The relative change direction and magnitude ratio of voltage and current phasors in the complex plane in multiphase lines are calculated. Based on the direction consistency clustering algorithm, the faults in the area and the disturbances outside the area are distinguished, and the interphase coupling migration behavior is identified accordingly.

[0014] By combining the harmonic amplification response characteristics and interphase coupling migration behavior, and by incorporating the attenuation or jump characteristics of the power frequency component within the time window, the transient characteristics of the power frequency are identified.

[0015] Optionally, S1 further includes classifying the main perturbation mechanism features and assigning a unique main perturbation mechanism type label to each type of feature.

[0016] Optionally, the multi-channel discriminant network is configured with independent processing channels corresponding to the main perturbation mechanism type label. Each channel, based on its corresponding mechanism type label, uses a built-in attention weight matrix to weight and focus the input phasor dynamic response sequence, calculates the contribution of each sequence component to the overall perturbation mode, and identifies one or more sequence components with the highest contribution as the phasor perturbation factor driving the response by comparing the contribution outputs across channels.

[0017] Optionally, S2 includes constructing a phasor factor mapping diagram by plotting time as the horizontal axis and phasor amplitude / phase angle as the vertical axis, and constructing the evolution trajectory of the identified dominant phasor perturbation factor in the complex plane.

[0018] Optionally, the factor stripping operation includes subtracting a comprehensive perturbation model consisting of the dominant phasor perturbation factor and its coupling components in the selected frequency band from the original phasor dynamic response sequence; this stripping operation generates a signal set that filters out the dominant perturbation components and highlights the essential change in the fundamental power frequency phase, i.e., the dominant purified phasor set.

[0019] Optionally, S3 includes calculating the first derivative of the power frequency phase angle of the voltage and current of each phase in the dominant purification phasor set with respect to time, identifying the moment when the phase angle derivative first exceeds a preset threshold of the phase angle change rate as the phase angle transition point of the phasor, and using this point as the fault initiation phase transition boundary.

[0020] Optionally, S3 further includes calculating the real-time phase difference and amplitude ratio between any two phase purification phase quantities. When the phase difference or amplitude ratio in the time series data undergoes a step jump, it is determined to be a phase difference mutation region. Based on the spatiotemporal correlation of multiple phase difference mutation regions, the cross-phase shift trend characterizing the fault nature in the phase propagation path is determined.

[0021] Optionally, the trend evolution direction is used to determine the fault type, specifically including: when the cross-phase offset trend is a phase transition of a single phase and the phase difference of other phases remains stable, it is identified as a single-phase ground fault; when the cross-phase offset trend is a near-synchronous phase transition and a continuous abrupt change in phase difference between two or more phases, it is identified as a phase-to-phase short circuit or cross-line fault.

[0022] The synchronization degree is used to locate the faulty phase, specifically including: by comparing the time series of each phase transition point with the phase angle jump direction, the phase with a transition point time difference less than a set tolerance and a phase angle jump direction that conforms to the fault electrical characteristics is determined as the faulty phase;

[0023] Based on the combined determination of the fault type and fault phase, the final fault identification result is output.

[0024] The beneficial effects of this invention are:

[0025] This invention constructs a progressive multi-time-window nested structure comprising millisecond-level short windows, cycle-level medium windows, and hundred-millisecond-level long windows. It performs multi-scale dynamic decomposition of voltage and current phasors in high-voltage lines, effectively covering the power fault response characteristics from transient disturbances and steady-state shifts to trend evolution. This design overcomes the contradiction between time-domain resolution and disturbance feature capture capability caused by traditional fixed-window or single-scale analysis. It enables early fault signs, abrupt change nodes, and subsequent inter-phase linkage behaviors to be fully identified and quantified at various time scales, providing a high-information-density dynamic phasor sequence foundation for subsequent high-precision mechanism classification and fault identification.

[0026] This invention proposes a mechanism type label-driven multi-channel discriminant network structure. Dedicated discriminant channels are configured for different types of perturbation mechanisms (power frequency transients, harmonic amplification, and interphase coupling), and an attention weighting mechanism is embedded within each channel to calculate the contribution index of each phasor component to the perturbation mode. This enables accurate identification of dominant perturbation factors with strong driving force and high information weight. Furthermore, by constructing a phasor factor mapping diagram and a frequency band coupling perturbation model, the dominant perturbation of the original phasor sequence is stripped and purified, effectively eliminating interference such as harmonics and asymmetric fluctuations. The purified phasor set, highlighting the intrinsic power frequency variation trend, is extracted, providing a clear physical basis and signal guarantee for subsequent phase transition identification.

[0027] This invention uses the abrupt change point of the first derivative of the phase angle with respect to time as the criterion for the transition boundary of the phasor response, and jointly analyzes the step-change region of the real-time phase difference and amplitude ratio to form a dual identification path for the fault initiation time and the cross-phase propagation trend. On this basis, by constructing a discrimination model of trend evolution direction + phase synchronization degree, it realizes the distinction between single-phase grounding and phase-to-phase short circuit / cross-line fault types. Furthermore, it introduces the time difference tolerance of the transition point and the consistency of the jump direction as the basis for determining the fault phase, ensuring the stability and physical consistency of the identification. The final output of the fault type and phase structured results not only has clear logical traceability, but also highly meets the needs of dispatch automation and protection system rapid response. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in this 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 for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the identification method flow according to an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the phase transition identification and fault output process according to an embodiment of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0032] like Figures 1-2 As shown, a high-voltage line fault identification method based on phasor analysis includes the following steps:

[0033] S1, Disturbance Mechanism Identification and Feature Classification: The voltage and current phasors collected during the operation of high-voltage lines are decomposed using a multi-time-window nested method to extract the dynamic response sequence of the phasors. Based on the characteristics of the disturbance frequency spectrum and the consistency of the response direction, the dynamic response sequence of the phasors is classified into mechanisms to identify the main disturbance mechanism features triggered by the fault and to construct a main disturbance mechanism type label. The main disturbance mechanism features include power frequency transients, harmonic amplification response and interphase coupling migration behavior, which are used for subsequent extraction of dominant factors.

[0034] S11, Phasor Dynamic Response Sequence Extraction: Extraction of voltage phasors from the synchronous phasor measurement unit (PMU) during the operation of high-voltage lines. With current phasor A progressive time window nesting decomposition is performed to generate multi-scale phasor dynamic response sequences. The specific steps are as follows:

[0035] S111. To capture response characteristics from different time scales, a progressive time window structure is established. Progressive time window nesting decomposition is a multi-scale time series analysis method. Its core is to analyze the same signal in parallel at multiple time scales, forming a hierarchical observation structure, enabling the system to have good perception capabilities between extremely short-term effects and longer-term trends. It emphasizes progressiveness and nesting; the progressive analysis window grows from short to long, and information gradually expands from local mutations to trend evolution; nesting emphasizes that the data structures within different time windows are mutually inclusive and complementary, not isolated. Specifically:

[0036] Millisecond short window Used to capture transient disturbances in the early stages of a fault; the initial transient changes in a power system usually occur within milliseconds, such as sudden changes in short-circuit current and voltage drops. The goal is to capture the initial high-speed disturbance characteristics to facilitate the identification of fault trigger points, instantaneous jumps and high-frequency components. The response time of protection relays is usually within the order of 10ms, so this window covers 0.25–1 power grid cycle, which meets the target of capturing high frequencies and transients.

[0037] Frequency-level medium window Used to identify steady-state offsets. In a 50Hz power grid system, one power frequency cycle is 20 ms. The goal is to capture the changes in steady-state response after a fault and identify whether steady-state current and voltage offsets occur. This time window covers a complete fundamental frequency cycle and can be used for steady-state feature identification such as single-cycle Fourier analysis and synchronous detection.

[0038] Millisecond-level long window Used to observe the evolution of response trends. The fault evolution process usually develops and is completed within hundreds of milliseconds. The goal is to observe trend evolution and potential low-to-medium frequency disturbances, such as harmonic amplification, energy transfer, and coupling feedback. PMU devices often sample at 10–50 frames per second, so a hundred-millisecond window provides a good balance, covering multiple periodic average characteristics while avoiding excessive smoothing.

[0039] S112. In each time window (include In the calculation of the magnitude and phase angle trajectories of voltage and current phasors:

[0040] ;

[0041] ;

[0042] in, Represents the modulus of a complex number. Represents the complex phase angle. These represent the time windows. Phasor data within, Indicates voltage phasors in the window The amplitude trajectory in Indicates the current phasor in the window The amplitude trajectory in Indicates voltage phasors in the window The phase angle trajectory in Indicates the current phasor in the window The phase angle trajectory in;

[0043] Within each time window, the amplitude trajectory (i.e., magnitude change) and phase angle trajectory (i.e., phase change) of the voltage phasor and current phasor are calculated respectively. These trajectories represent the changes in signal strength and direction over time. Through this operation, we can not only know whether the current / voltage has become stronger, but also how they rotate or jump in the complex plane. This is crucial for identifying certain types of faults (interphase coupling, harmonic amplification).

[0044] S113. The amplitude and phase trajectories extracted from different time windows are packaged and combined into a unified response sequence set. Each item contains observations from short, medium, and long window perspectives, forming a dynamic response view with rich temporal resolution and distinct scale levels. This sequence will become a key input for subsequent perturbation mechanism identification and dominant factor extraction. In other words, by summarizing the results from all time windows, a phasor dynamic response sequence is constructed, represented as: .

[0045] S12, Identification and classification of main perturbation mechanism features: For the above phasor dynamic response sequences... Frequency domain and directional consistency analysis is performed to identify the main fault disturbance mechanism, specifically including:

[0046] S121. Perturbation Frequency Spectrum Characteristic Analysis (Identifying Harmonic Amplification Response): Apply Fast Fourier Transform (FFT) to the phasor amplitude trajectory within each time window:

[0047] ; Indicates the voltage phasor magnitude at the th Fast Fourier Transform results within the time window. Indicates the magnitude of the current phasor in the th order. Fast Fourier Transform results within the time window. For frequency variables;

[0048] Calculate the energy percentage of a specific frequency band to determine harmonic amplification behavior:

[0049] ;

[0050] in, It is a set of harmonic frequency bands (including the second and third harmonics and interharmonic frequency ranges). Hz is the fundamental frequency. Represents the frequency domain spectral value. Indicates the first Harmonic frequency band energy integral within the time window Indicates the first Fundamental frequency energy within the time window It indicates the proportion of harmonic energy and is used to determine the degree of harmonic amplification response.

[0051] Disturbance frequency spectrum analysis is used to determine whether harmonic amplification exists. A Fast Fourier Transform (FFT) is used to transform the phasor amplitude over time in the frequency domain, revealing the energy distribution at different frequencies (e.g., fundamental, second harmonic, third harmonic, etc.). By calculating the ratio of the total energy of the harmonic bands to the fundamental energy, it can be determined whether the current fault has caused significant harmonic enhancement. If the proportion of harmonic energy is high, it indicates harmonic amplification in the system, which may be related to phenomena such as electric arcing, nonlinear loads, or resonance.

[0052] S122. Response Direction Consistency Analysis (Identifying Interphase Coupling Migration Behavior): Constructing change vectors for multiphase voltage and current phasors in the complex plane: ; These are the voltage phasors for phase a and phase b, respectively. These are the current phasors for phase a and phase b, respectively. Let be the difference vector of the voltage phasors of phases a and b, indicating the direction of the phase-to-phase voltage change. Let be the difference vector of the phasors of the currents in phases a and b, indicating the direction of the interphase current change.

[0053] For a three-phase system (phases A, B, and C), phase combinations such as AB, BC, and CA can be constructed. For each combination, the direction cosine value at each moment (i.e., the consistency of the phasor change direction) is calculated. That is, the direction cosine of the change between different phases is calculated. Taking AB as an example, the calculation is as follows: ;in, This represents the real part operator. Indicates complex conjugation. The cosine similarity index, representing the direction of voltage and current change between phases a and b, is used to measure directional consistency. The value falls in the interval [−1, 1], and the closer it is to 1, the more consistent the direction. Using a period of time as the observation window, the directional consistency trajectory between each pair of phases can be represented as a time series vector, and then concatenated into a unified feature matrix. The directional consistency feature vector corresponding to each phase combination is regarded as a data point, forming a clustering input space. For example, each phase combination is treated as a sample, and the sample characteristics are its time series of directional consistency (mean, standard deviation) within the observation window. By selecting appropriate DBSCAN (density clustering), high-consistency clusters (i.e., with strong interphase coupling) and low-consistency or isolated perturbations can be automatically identified. The clustering results are divided into several categories. If two or more phase combinations are classified into the same category and the directional consistency index is significantly high, it indicates that there is a common phasor perturbation trend among these phases. This is the criterion for interphase coupling migration behavior, that is, phasor perturbations exhibit synchronicity and coupling among multiple phases. When a persistent and highly consistent clustering pattern is found among certain phase combinations, it can be determined that the current fault response belongs to the interphase coupling migration type perturbation mechanism, and an interphase coupling label can be assigned accordingly.

[0054] The response direction consistency analysis determines whether interphase coupling migration exists. This part focuses on whether the voltage and current change directions are consistent between different phases (e.g., phase A, phase B). By constructing complex change vectors (voltage difference and current difference) between two phases and calculating their direction cosines, the existence of coupling behavior can be determined. If the directions of multiple phases are highly consistent, it indicates that the fault impact is not limited to a single phase and there may be cross-phase transmission or resonant coupling phenomena. In this case, it can be identified as interphase coupling migration behavior. This analysis ultimately uses a clustering algorithm to automatically classify all direction cosines and identify whether there are clustered coupling phenomena.

[0055] S123. Power Frequency Transient Identification and Tag Construction: Analyzing abrupt changes in amplitude and phase jumps at power frequency components: If the power frequency component shows a significant jump or rapid decay, accompanied by high... If the phase-to-phase coupling is strong, then the presence of power frequency transient characteristics is determined. This step involves analyzing the behavior of the power frequency (50Hz) components to see if there are abrupt amplitude changes (jumps), significant phase angle changes, or attenuation. If these changes exist, and harmonic enhancement or directional coupling phenomena are also present, then the presence of power frequency transient characteristics is determined, indicating that the fault has strong disturbance characteristics. This is specifically divided into two steps:

[0056] Step 1: Analyze the amplitude and phase angle changes of the power frequency components: within each time window Inside, the system will track the amplitude trajectory of the voltage phasor. Observe it, paying particular attention to its changing trend on the time scale corresponding to the power frequency, using the above-mentioned... To calculate a short time interval The instantaneous amplitude change within a certain period; if this difference continuously changes drastically (jumps) or decreases rapidly (attenuates) over a period of time, it indicates that an abnormal disturbance has occurred in the power frequency component of the voltage amplitude; at the same time, if a sudden change in phase angle is also observed (i.e., an instantaneous phase jump), it further verifies the existence of a serious disturbance. This situation often corresponds to the following in power systems:

[0057] The short circuit fault has just occurred or has just been cleared;

[0058] Transient processes caused by power switching, load surges, etc.

[0059] Step 2: Comprehensive assessment of whether it constitutes a power frequency transient characteristic: relying solely on changes in power frequency amplitude may not be sufficient for accurate determination, therefore the scheme also requires two auxiliary conditions to be met:

[0060] High proportion of harmonic energy ( High): This indicates that a large number of harmonic components are generated during the mutation process;

[0061] Strong interphase coupling indicates that the disturbance is not isolated in a single phase, but rather has cross-phase propagation characteristics.

[0062] Only when these three factors appear simultaneously or corroborate each other can the current event be considered to have the characteristics of "power frequency transients". This is a composite disturbance characteristic, reflecting that the fault involves not only drastic energy changes, but also frequency domain complexity and spatial coupling.

[0063] S13, Finally, based on the combination patterns of the three types of characteristics (harmonic amplification response, interphase coupling migration, and power frequency transient), a primary disturbance mechanism type label is assigned. The system assigns a primary disturbance mechanism type label to each type of fault disturbance, including power frequency transient, harmonic amplification, and interphase coupling. This label will serve as an important guiding basis for subsequent steps and is represented as follows:

[0064] .

[0065] S2, Dominant Factor Extraction and Phasor Purification: The main perturbation mechanism features are input into a multi-channel discrimination network. Under the guidance of the main perturbation mechanism classification label, the phasor perturbation factors that drive the response are identified. A phasor factor mapping map is constructed and a factor stripping operation is performed to generate a dominant purified phasor set. The dominant purified phasor set is used for subsequent phase transition identification.

[0066] S21, Phasor perturbation factor identification under a multi-channel discriminant network: identifying the main perturbation mechanism features and their corresponding type labels. Input to multi-channel discrimination network Identify the phasor perturbation factors that play a dominant role in the perturbation response.

[0067] S211, the multi-channel discriminant network structure is defined as follows:

[0068] ;

[0069] in, For the corresponding main disturbance mechanism label Each channel has an independent discrimination channel, and each channel independently processes the input phasor dynamic response sequence. .

[0070] S211 uses the identified main perturbation mechanism type labels as prior information to guide a specially designed multi-channel discriminant network to identify which phasor components play a key role in the fault response. The most crucial structure is the independent discriminant channel. The multi-channel discriminant network is a neural network architecture containing multiple parallel but independent sub-modules, i.e., channels, each dedicated to processing a specific category of data or signal features. In this invention, each main perturbation mechanism type label is considered to have different perturbation patterns and feature focuses. Therefore, to better distinguish the key perturbation factors under these mechanism types, an independent processing channel is designed for each label type, specifically learning the phasor perturbation features under that mechanism type. The independent discriminant channel is an independent sub-network branch within the entire discriminant network corresponding to a specific perturbation mechanism label. Each channel possesses the following characteristics:

[0071] Input sharing, weight independence: Each channel receives the same input (i.e., the entire phasor dynamic response sequence), but its network parameters (weights, biases, attention mechanisms) are specifically trained for the corresponding mechanism type and are not shared with other channels.

[0072] Targeted Feature Focus Mechanism: Each channel contains an attention weighting module to identify which phasor components are most important for that type of perturbation. For example, the harmonic amplification channel focuses more on frequency components, the interphase coupling channel emphasizes interphase direction changes, and the power frequency transient channel focuses on phase angle behavior at abrupt change points. Multiple channels run simultaneously, each outputting its focus on the perturbation mode (i.e., perturbation factor contribution assessment), but they do not overlap or influence each other, ensuring the independence and professionalism of the analysis results for each mechanism.

[0073] The diversity of fault disturbance mechanisms means that the key phasor factors of different types of disturbances are often distributed across different dimensions or feature combinations. Using a single model to identify all disturbance types can easily lead to unclear model focus, inability to distinguish different mechanisms, and the average masking of key disturbance factors. By introducing independent discrimination channels, the discrimination process for each disturbance mechanism becomes a specialized task, significantly improving the accuracy of causal attribution. Each channel outputs a set of "contribution" evaluation values ​​for the disturbance factors, used to determine which phasor component is considered most decisive. By comparing the evaluation results of these channels, the most critical dominant phasor disturbance factors can be accurately extracted, providing a cleaner and more representative signal for subsequent phase transition identification.

[0074] S212, Attention-weighted focusing mechanism: Each channel contains an attention weight matrix. This is used to measure the degree of attention each component in the input sequence receives to the perturbation response. The channel, attention-weighted representation is as follows: ;in, This represents element-wise multiplication. This is the weighted sequence representation. Indicates the number of time steps. The feature dimensions include combinations of phase angle, amplitude, and multiphasors. The core of this section explains how each independent discrimination channel focuses on the importance of different phasor components. Its key technical mechanism is the introduction of attention-weighted focusing. In the dynamic response sequence composed of voltage and current phasors, not every moment or feature has the same importance to the disturbance. For example, if the voltage of a phase jumps instantaneously at the initial stage of a fault, the amplitude change during this period is particularly important. Or, if the phase angles of multiple phases fluctuate synchronously within a certain period, it may also indicate potential coupled disturbances. Therefore, it is necessary to identify which time points and which components are most critical to the fault response; this is the role of the attention mechanism. An attention weight matrix is ​​introduced into each channel, denoted as... In this matrix, rows represent different time steps (i.e., each sampling moment), columns represent different feature dimensions (phasor magnitude, phase angle, phase difference), and the value of each matrix element represents the degree of attention of the corresponding time-feature in that channel; the larger the value, the more important it is. The original input sequence is... matrix It records the evolution of all phasor features at each time point, and this sequence is compared with the attention weight matrix. Perform element-wise multiplication to obtain a new weighted sequence. The essence of this operation is to preserve the original information of important components (high-weight positions) and attenuate (or even erase) unimportant components (low-weight positions) to obtain a focused phasor sequence that retains only the part most relevant to the current channel (i.e., the corresponding perturbation mechanism).

[0075] S213, Calculation and Extraction of Phasor Perturbation Factor Contribution: In each channel, calculate the contribution index of each sequence component (including the amplitude trajectory under a certain phase and time window) to the output perturbation classification decision. :

[0076] ;in, Let be the classification loss function of the network. For the first The input component in the th ... The weighted values ​​in the channel are compared by comprehensively assessing the contribution of all components, and one or more sequence components corresponding to the maximum value are selected as the set of dominant phasor perturbation factors. These dominant factors are key fragments selected from the complete response sequence, i.e. .

[0077] The above content mainly focuses on identifying the phasor components that play a key role in the fault perturbation mechanism from a large number of phasor dynamic response sequences, relying on a quantization mechanism, namely, the calculation of the contribution of perturbation factors. Previously, a multi-channel discriminant network was used to classify and judge the fault perturbation mechanism. Each channel corresponds to one perturbation mechanism, and each channel uses an attention mechanism to weight the input sequence, resulting in weighted inputs with different focus angles for different channels. However, the problem lies in identifying which of these input components are the core components that truly drive the network to make a judgment. These need to be identified, i.e., the so-called dominant perturbation factors. To solve this problem, this section introduces the concept of a contribution index, which can be simply understood as: how much influence each input component has on the current classification result. This influence is measured by the sensitivity (gradient) to the classification loss function. Specifically: if a slight change in a component leads to a significant change in the classification result, it is considered crucial and has a high contribution; if a change in a component has almost no impact on the classification result, it is considered unimportant and has a low contribution.

[0078] In the network, each channel performs a classification calculation on the weighted input (the attention-weighted phasor sequence). At this point, we can calculate the partial derivative of the network's loss function, i.e., the classification error, with respect to each input component. The loss function measures how much the network misclassifies; the partial derivative measures how much the loss function changes if this component changes slightly. This partial derivative value is the so-called contribution index. If the partial derivative with respect to a certain input component is large, it indicates that it has a significant impact on the classification judgment and is a potential key perturbation factor. Each channel calculates the contribution of all input components, and then compares these contributions uniformly, both across channels and within a single channel, selecting the components with the largest contribution values. These are considered to be the phasor components that elicit the strongest response under the current dominant perturbation mechanism and play a decisive role in fault identification. These selected components are the final dominant perturbation factors, denoted as a set, for subsequent stripping and purification to more clearly identify the true power frequency phase anomalies.

[0079] S22, Phasor factor mapping construction and factor stripping operation:

[0080] S221, Phasor Factor Mapping Construction: For the identified set of dominant perturbation factors Its evolution trajectory in time and the complex plane is plotted as a phasor factor mapping. Its structure is as follows:

[0081] ;in, Let them represent the real and imaginary parts of a complex number, respectively. As the dominant factor at time The complex-valued phasors are used to visually reflect the dynamic propagation mode and energy focusing region of the dominant disturbance in the complex plane.

[0082] Phasor factor mapping diagram This indicates at each point in time. Record the real part of the dominant perturbation factor phasor and the virtual part This is equivalent to plotting trajectory points in a complex plane, where each point is determined by the complex-valued phasor at a certain moment in the time series. If we understand this graph as a regular two-dimensional graph, where the horizontal axis represents time... The vertical axis uses complex coordinates (real and imaginary parts) to represent the complex values ​​of phasors, and plots the phasor magnitudes (i.e.,...) on the vertical axis. ) or phase angle (i.e. This allows for a clearer representation of the phasor's evolution over time, in a format where time is the horizontal axis and amplitude / phase angle is the vertical axis.

[0083] Section S221 graphically represents the evolution of the identified dominant disturbance factors over time, particularly their trajectories in the complex plane. This helps observe how fault signals evolve from weak disturbances to significant anomalies, and how disturbance energy concentrates or diffuses. The phasor factor mapping is a graph that displays the disturbance evolution in the form of complex trajectories. Each point in the graph represents the complex phasor of the dominant disturbance factor at a certain moment. The phasor factor mapping allows us to visually observe the dynamic behavior of the disturbance factors in the complex plane. For example:

[0084] If the trajectory exhibits a spiral, stretching, or abrupt change shape, it indicates that the signal has been subjected to non-steady-state disturbances.

[0085] If the trajectories cluster in a certain area, it may reflect a local concentration of energy;

[0086] If the trajectory deviates significantly from the normal circular fluctuation trajectory, it suggests that there may be problems such as asymmetric faults, coupling propagation, or harmonic amplification.

[0087] Traditionally, the phasor-time curves we often draw are:

[0088] Horizontal axis: Time;

[0089] Vertical axis: amplitude or phase angle.

[0090] The phasor factor mapping diagram here is:

[0091] At any given moment, the complex phasor forms a point on the complex plane;

[0092] Multiple moments linked together form a dynamic trajectory.

[0093] This representation method preserves both amplitude and phase angle information, making it more suitable for analyzing complex disturbance behaviors such as multiphase coupling and nonlinear responses.

[0094] S222, Factor Stripping Operation (Disturbance Purification): In actual power systems, signals during faults are often mixed with power frequency components and various strong nonlinear disturbances, including harmonic amplification, phase coupling, and abrupt jumps. If we directly analyze these mixed signals, we may be misled by the interference components, leading to inaccurate judgments of the fault initiation point and fault phase. Therefore, it is necessary to first identify and strip these dominant disturbance factors from the original signal, leaving only the purified power frequency response signal that reflects the true fault trajectory. To extract the true fundamental response characteristics, the aforementioned disturbance factors need to be stripped from the original response sequence.

[0095] Before stripping, a perturbation model containing the dominant perturbation factor must first be constructed. This model is not arbitrarily subtracted, but rather combined:

[0096] The dominant disturbance factor extracted earlier is taken as the source signal;

[0097] Observe the spectral response of these factors in a specific frequency band;

[0098] These frequency band components are projected and weighted to represent their contribution to the overall fault disturbance;

[0099] The contributions of all these frequency bands are then weighted and summed to form a comprehensive disturbance model.

[0100] This model can be viewed as the sum of all the interference waveforms generated by these dominant factors in the fault signal.

[0101] Specifically, this includes: constructing a comprehensive perturbation model consisting of the dominant perturbation factor and its coupled components in a specific frequency band. :

[0102] This formula constructs a comprehensive perturbation model to simulate the influence intensity and propagation behavior of the dominant perturbation factor at multiple key frequencies. Essentially, it combines perturbation components from different frequencies in proportion to form a subtractable interference model, thereby purifying the original signal. The coupling strength coefficient (calculated from the spectral response of the dominant factor) is... For the disturbance factor at frequency point phasor projection components on It is a set of harmonic frequency bands. This is the fundamental frequency.

[0103] This indicates that the frequency components we are considering include both the fundamental frequency and a specific set of harmonic frequency bands. A specific frequency band refers to the frequency band other than the fundamental frequency. Beyond the normal operating components, special attention is paid to harmonic frequency bands. These bands are precisely where harmonic amplification, nonlinear disturbances, and fault resonance are concentrated. Therefore, they are extracted and coupled with the dominant disturbance factor to form a model for subsequent separation. Examples include the fundamental frequency (50Hz), harmonic frequencies (100Hz, 150Hz), interharmonic frequencies, and non-integer multiples of the frequency range. With the integrated disturbance model, the purified dominant phasor set can be obtained by subtracting it from the original phasor response sequence. : The purification results filtered out the masking and interference of the dominant disturbance on the fundamental wave signal, highlighting the true trend of power frequency phase change, and providing a stable foundation for subsequent phase transition point identification.

[0104] The purification process removes interference, improves the accuracy of subsequent analysis, and preserves key phase evolution characteristics, providing a clear and stable signal basis for the next step of identifying phase transition points and fault initiation boundaries.

[0105] S3, Phase transition identification and fault output: Perform phase evolution analysis on the dominant purification phasor set, identify the phase angle transition point of the phasor and the abrupt change region of the phase difference, determine the fault initiation phase transition boundary and cross-phase offset trend, and output the fault to which the phase and type identification result belongs based on the trend evolution direction and the degree of synchronization.

[0106] S31, Identification of phase transition boundary and cross-phase offset trend:

[0107] S311, Identify phasor phase angle transition points: For the dominant purification phasor set Each phase voltage phasor or current phasor Calculate the rate of change of its power frequency phase angle with time:

[0108] ;

[0109] or: ;

[0110] At a certain moment satisfy: Then, this moment is identified as the phasor phase angle transition point and recorded as the fault initiation phase transition boundary of that phase, where, Indicates the three-phase number. Represents the complex phase angle. Represents the phase angle derivative. The preset threshold representing the rate of change of phase angle is used to determine whether the derivative of phase angle is large enough to indicate an abnormal change. Only when the rate of change of phase angle exceeds this value is it considered not a normal fluctuation, but the beginning of a possible fault. Based on the analysis of a large amount of power system phasor data, the threshold is set to [30, 100] rad / s. For sensitive identification, it is recommended to set it to a lower value, 30–50 rad / s; for robust identification, i.e., only identifying severe disturbances, it can be set to 80–100 rad / s; or it can be adaptively adjusted by using a sliding window mean + standard deviation dynamic threshold setting method.

[0111] The purpose of the above is to automatically identify the starting point of a fault within the dominant set of phasors to be cleaned, that is, the point at which the voltage or current signal of a certain phase begins to change significantly. This point in time is called the phase transition point, which represents the break point where the system transitions from normal operation to a fault state.

[0112] Specifically, S311 above calculates the change of the phase angle of each phase (A, B, C) voltage or current signal in the complex plane over time. This phase angle reflects the rotational state of the phasor on the unit circle and directly reflects the phase characteristics of the voltage or current in the power system. Differentiating this phase angle with respect to time yields its rate of change, i.e., how much the phase angle changes per unit time, called the phase angle derivative or phase rotation rate. If the phase angle derivative at a certain moment, i.e., the rate of change, suddenly exceeds a set threshold, this moment is considered the beginning of a rapid phase change, i.e., the phasor phase angle transition point. This transition point is marked as the fault initiation boundary of that phase. The transition points of multiple phases can also be used for subsequent identification of the fault phase (which phase it occurs in) and type (single-phase or multi-phase).

[0113] S312, Identifying abrupt changes in phase-to-phase differences: This identifies whether cascading disturbances have occurred in the three-phase power grid system, i.e., whether the fault signal is not limited to one phase but has propagated across multiple phases. To achieve this, continuous tracking and analysis of the phase angle difference and amplitude ratio between two phases is performed to determine whether a significant abrupt change (step jump) has occurred between the two phases within a certain time period. Details are as follows:

[0114] For any two phases Calculate the following two indicators:

[0115] Extracting real-time phase difference: ; This represents the voltage phasors of phases A / B / C. The phase angle is the difference between the two phases. If this difference suddenly changes within a short period of time, it indicates that the phase synchronization of the two phases has been disrupted, and there may be coupling or transfer.

[0116] Calculate the real-time amplitude ratio: This ratio reflects the degree of voltage amplitude imbalance. A jump indicates abnormal amplitude fluctuations in one phase and may also foreshadow cross-phase interference.

[0117] Judgment: Within a small time interval, when the following is detected in the time series: or In a certain short period of time A step jump occurs within the system, that is:

[0118] ;

[0119] or: ;

[0120] Then the region at that time point is determined to be a region of abrupt changes in phase difference.

[0121] in, This is the phase difference jump threshold, used to determine whether the phase difference between two phases has undergone an abnormal abrupt change. Under normal operating conditions, the phase angle difference between phases will fluctuate slightly around a fixed value. During a fault, the phase of one phase may jump by tens of degrees instantaneously, causing the phase difference between the two phases to suddenly increase or decrease. The unit is radians (rad), which is approximately 30° ≈ 0.523 rad, with a value range of [0.3, 0.7] rad. Alternatively, a dynamic threshold setting using a sliding window mean + standard deviation can be used for adaptive adjustment. The amplitude ratio jump threshold is used to detect whether the voltage amplitude ratio of two phases suddenly becomes unbalanced. Under normal circumstances, the three-phase voltage amplitudes are close, and the ratio is in the range of 0.95–1.05. During a fault, the voltage of one phase suddenly drops or rises, causing the ratio to deviate significantly. The value range is [0.1, 0.3]. Alternatively, a sliding window mean + standard deviation dynamic threshold setting method can be used for adaptive adjustment.

[0122] The system tracks abrupt changes in multiple phase combinations (e.g., AB, BC, CA) and determines whether these changes exhibit temporal continuity and spatial path logic. If multiple phase abrupt changes form a chain, a cross-phase shift trend is identified, indicating that the fault signal may be a process of gradual transmission from one phase to other phases. In other words, by combining the abrupt change regions of multiple phase pairs, a distribution map of these abrupt changes in time and along the phase paths is constructed. Based on their concentration and evolution direction, a cross-phase shift trend is determined, i.e., the synchronous disturbance characteristic of a fault signal originating in one phase and gradually propagating to other phases.

[0123] S32, Fault type and phase determination:

[0124] S321, Fault type determination, based on trend evolution direction:

[0125] If only one phase (e.g., phase A) experiences a phase transition, while the phase difference between the other two phases remains stable throughout the process, this indicates that the fault is localized, affecting only one phase while the other two phases continue to operate normally without any obvious abnormalities. Therefore, the fault type is determined to be a single-phase ground fault.

[0126] If two or more phases (such as phases B and C) undergo phase transitions almost simultaneously, accompanied by continuous abrupt changes in phase-to-phase differences, this indicates that the fault is not confined to one phase but rather involves coupling propagation between phases. This could be due to a two-phase short circuit, a jumper in one phase affecting the electromagnetic characteristics of another phase, or two of the three phases simultaneously experiencing anomalies. Therefore, the fault type is determined to be either a phase-to-phase short circuit fault or a cross-line fault.

[0127] S322, Fault Phase Identification, aims to determine which phase(s) has actually failed, i.e., to accurately pinpoint the faulty phase. In high-voltage transmission systems, when a fault occurs, not all phases will be abnormal simultaneously. It could be that phase A is independently grounded; or phases B and C are short-circuited; or all three phases (A, B, and C) may simultaneously change phase. Therefore, one cannot simply look at the abnormal amplitude or phase angle of a single phase, but must comprehensively judge based on the synchronicity of the phase changes, i.e., based on the degree of synchronicity analysis. In the previous steps, the phasor phase angle abrupt change of each phase, i.e., the transition point, has been identified. For all phases... Mark the time of these turning points as Furthermore, the phase angle transition direction is also needed. For each phasor signal, observe whether its phase angle increases or decreases at the transition point. The phase angle transition direction is represented as follows: Phase angle transition time and the direction of phase angle jump Perform time synchronization and transition direction matching. If two or more phases satisfy the following conditions:

[0128] These phases are then considered to be fault-involved phases.

[0129] The above formula means that if the transition times of two or more phases are very close, i.e., the time difference is less than a certain set tolerance; and their phase angle jump directions are consistent or close; then it indicates that they are phases that participated in the fault together, belonging to the coupled response under the same disturbance. The sign function represents the direction of the transition. Indicates the amplitude of the phase angle jump. The maximum tolerance time difference is set, with values ​​ranging from [10ms, 40ms]. For ordinary PMU systems, it is set to 20–30ms, while for high-precision synchronization systems, it is reduced to 10–15ms. If greater response fluctuations or inter-network communication delays are tolerable, it can be appropriately relaxed to 40ms.

[0130] Finally, all phases that satisfy both time synchronization and consistent jump direction are marked as the phase to which the fault belongs. For example:

[0131] A phase sudden change → the faulty phase is A;

[0132] B and C both mutate simultaneously → the fault phases are BC;

[0133] Three-phase synchronous sudden change → faulty phases are A, B, and C.

[0134] S33, combined with:

[0135] Fault type identification (single-phase grounding / phase-to-phase short circuit);

[0136] And fault phase location (A / B / C / AB / BC / AC, etc.);

[0137] The system outputs structured fault identification results:

[0138] Fault Result={Type:"Single-phase grounding",Phase:"A"}; indicates that the fault is identified as a single-phase grounding fault, occurring in phase A, meaning that only the voltage / current of phase A is abnormal, while phases B and C remain normal.

[0139] Alternatively: FaultResult={Type:"Phase-to-Phase Short Circuit",Phase:"BC"}; indicates that the fault type is identified as phase-to-phase short circuit, that is, an electrical short circuit has occurred between two phases. The fault involves phases B and C, indicating that the phase angle and amplitude of these two phases have undergone abnormal changes that are close to synchronization.

[0140] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0141] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A high-voltage line fault identification method based on phasor analysis, characterized in that, Includes the following steps: S1. The voltage and current phasors collected during the operation of the high-voltage line are decomposed into multiple time windows to extract the dynamic response sequence of the phasors. Based on the frequency spectrum characteristics of the disturbance and the consistency of the response direction, the dynamic response sequence of the phasors is classified into mechanisms to identify the main disturbance mechanism characteristics triggered by the fault and to construct a main disturbance mechanism type label. The main disturbance mechanism characteristics include power frequency transient, harmonic amplification response and interphase coupling migration behavior. S2, input the main perturbation mechanism features into a multi-channel discrimination network, identify the phasor perturbation factors that drive the response under the guidance of the main perturbation mechanism classification label, construct a phasor factor mapping map and perform factor stripping operation to generate a set of dominant purified phasors; S3, perform phase evolution analysis on the dominant purification phasor set, identify the phase angle transition point of the phasor and the abrupt change region of the phase difference, determine the fault initiation phase transition boundary and cross-phase offset trend, and output the fault to which the phase and type are identified based on the trend evolution direction and the degree of synchronization.

2. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, The extracted phasor dynamic response sequence specifically includes: A progressive time window structure is used to perform multi-scale analysis on voltage and current phasor data collected by the synchronous phasor measurement device. The progressive time window structure includes a millisecond-level short window for capturing transient processes, a cycle-level medium window for identifying steady-state shifts, and a hundred-millisecond-level long window for observing trend evolution. By parallel calculation of the phasor amplitude and phase angle change trajectories within each time window, a phasor dynamic response sequence that can characterize the dynamic response of the entire fault process is generated.

3. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, In S1: The phasor dynamic response sequence is subjected to a fast Fourier transform to extract its perturbation frequency spectrum characteristics. By calculating the energy integral and fundamental energy ratio of a specific frequency band, the harmonic amplification response characteristics are identified. The specific frequency band includes the second harmonic, the third harmonic, and interharmonics. The relative change direction and magnitude ratio of voltage and current phasors in the complex plane in multiphase lines are calculated. Based on the direction consistency clustering algorithm, the faults in the area and the disturbances outside the area are distinguished, and the interphase coupling migration behavior is identified accordingly. By combining the harmonic amplification response characteristics and interphase coupling migration behavior, and by incorporating the attenuation or jump characteristics of the power frequency component within the time window, the transient characteristics of the power frequency are identified.

4. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, The S1 further includes classifying the main perturbation mechanism features and assigning a unique main perturbation mechanism type label to each type of feature.

5. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, The multi-channel discriminant network is configured with independent processing channels corresponding to the main perturbation mechanism type label. Each channel, based on its corresponding mechanism type label, uses a built-in attention weight matrix to weight and focus the input phasor dynamic response sequence, calculates the contribution of each sequence component to the overall perturbation mode, and identifies one or more sequence components with the highest contribution as the phasor perturbation factor driving the response by comparing the contribution outputs across channels.

6. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, S2 includes constructing a phasor factor mapping diagram by plotting time as the horizontal axis and phasor amplitude / phase angle as the vertical axis, and constructing the evolution trajectory of the identified dominant phasor perturbation factor in the complex plane.

7. The high-voltage line fault identification method based on phasor analysis according to claim 6, characterized in that, The factor stripping operation is achieved by subtracting a comprehensive perturbation model consisting of the dominant phasor perturbation factor and its coupling components in the selected frequency band from the original phasor dynamic response sequence; this stripping operation generates a signal set that filters out the dominant perturbation components and highlights the essential change in the fundamental power frequency phase, namely the dominant purified phasor set.

8. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, S3 includes calculating the first derivative of the power frequency phase angle of the voltage and current of each phase in the dominant purification phasor set with respect to time, identifying the moment when the phase angle derivative first exceeds a preset threshold of the phase angle change rate as the phase angle transition point of the phasor, and using this point as the fault initiation phase transition boundary.

9. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, S3 further includes calculating the real-time phase difference and amplitude ratio between any two phase purification phase quantities. When the phase difference or amplitude ratio in the time series data undergoes a step jump, it is determined to be a phase difference mutation region. Based on the spatiotemporal correlation of multiple phase difference mutation regions, the cross-phase shift trend characterizing the fault nature in the phase propagation path is determined.

10. The high-voltage line fault identification method based on phasor analysis according to claim 1, characterized in that, The trend evolution direction is used to determine the fault type, specifically including: when the cross-phase offset trend is a phase transition of a single phase and the phase difference of other phases remains stable, it is identified as a single-phase ground fault; when the cross-phase offset trend is a near-synchronous phase transition and a continuous abrupt change in phase difference between two or more phases, it is identified as a phase-to-phase short circuit or cross-line fault. The synchronization degree is used to locate the faulty phase, specifically including: by comparing the time series of each phase transition point with the phase angle jump direction, the phase with a transition point time difference less than a set tolerance and a phase angle jump direction that conforms to the fault electrical characteristics is determined as the faulty phase; Based on the combined determination of the fault type and fault phase, the final fault identification result is output.

Images