Electric power operation leakage protection hierarchical configuration method and system based on risk assessment

Abstract

The invention provides a risk assessment-based electric power operation leakage protection hierarchical configuration method and system. The method comprises the following steps of: obtaining zero-sequence current information of a plurality of electric power loops; acquiring an electrical connection relation and a current distribution proportion among the plurality of power loops; according to the zero-sequence current information, the electrical connection relation and the current distribution proportion, reconstructing the real total electric leakage amount of the plurality of power loops; judging whether an electric leakage risk exists or not according to the real total electric leakage amount; according to the characteristics of the zero-sequence current information, distinguishing currents from different sources to avoid a protection action caused by non-fault current superposition; and when it is judged that the electric leakage risk exists, triggering a corresponding protection action. The real electric leakage risk in a complex electric power environment can be effectively identified, the protection failure or misoperation of a traditional protection system caused by the dynamic change of electrical connection is avoided, and the safety of electric power operation is remarkably improved.

Description

Technical Field

[0001] This application relates to the field of power safety technology, and more specifically, to a method and system for hierarchical configuration of leakage current protection for power operations based on risk assessment. Background Technology

[0002] In large-scale industrial production environments, ensuring the safety of electrical work is paramount, and risk-assessed residual current devices (RCDs) are typically deployed. However, in complex environments with multiple work areas and temporary power circuits that are independent yet potentially interconnected due to temporary operations by on-site personnel, traditional RCDs face significant challenges. For example, when workers temporarily run power lines between different work areas or borrow power from other areas to respond to emergencies or expedite progress, these temporary electrical connections subtly alter the structure of the entire power network and the current flow path.

[0003] Traditional residual current circuit (RCC) systems are typically designed based on the assumption that each power circuit is independent, making them ill-suited to dynamic changes. This can lead to situations where, in the event of a leakage current, the current is diverted, causing the actual current detected by the local protection device to fall below its tripping threshold, resulting in protection failure and prolonged exposure of workers to dangerous contact voltages. For example, in a large industrial production facility, if an old handheld electric grinder in work area A generates a 22mA leakage current to ground, and the 30mA RCC in work area A fails to trip in time, workers will face a serious risk of electric shock.

[0004] On the other hand, such complex electrical connections can also lead to the superposition of weak leakage currents from multiple circuits, causing malfunctions in protection devices in non-faulty areas. For example, if the leakage current from the grinding machine in work area A and the common-mode leakage current generated by the precision diagnostic equipment in work area B are superimposed at the leakage current protection device in work area B, and their directions are consistent, it may cause the more sensitive 10mA leakage current protection device in work area B to malfunction, resulting in the interruption of critical operations, economic losses, and project delays. It may even introduce greater safety hazards because operators attempt to bypass the protection device. In this situation, static, localized risk assessment and graded protection strategies are completely inadequate to cope with systemic changes in electrical connections caused by temporary operations, leading to the failure or malfunction of the protection system at critical moments, seriously threatening operational safety. Summary of the Invention

[0005] This application discloses a risk assessment-based method for hierarchical configuration of leakage current protection for power operations, in order to solve at least one of the above-mentioned technical problems in the prior art.

[0006] Firstly, this application discloses a risk-assessment-based method for hierarchical configuration of leakage protection for electrical operations, including: Obtain zero-sequence current information from multiple power circuits; Obtain the electrical connection relationships and current distribution ratios among these multiple power circuits; Based on the zero-sequence current information, the electrical connection relationship, and the current distribution ratio, the actual total leakage current of the multiple power circuits is reconstructed. Based on the actual total leakage current, determine whether there is a risk of leakage current. Based on the characteristics of this zero-sequence current information, currents from different sources are distinguished to avoid protection actions caused by the superposition of non-faulty currents; and When a risk of leakage is detected, the corresponding protection action is triggered.

[0007] Secondly, this application also discloses a risk assessment-based method system for hierarchical configuration of leakage current protection in power operations, the system comprising: The zero-sequence current information acquisition module is used to acquire zero-sequence current information of multiple power circuits. The electrical connection relationship acquisition module is used to acquire the electrical connection relationship and current distribution ratio between the multiple power circuits; The leakage current total reconstruction module is used to reconstruct the actual leakage current total of the multiple power circuits based on the zero-sequence current information, the electrical connection relationship, and the current distribution ratio. The leakage risk assessment module is used to determine whether there is a leakage risk based on the actual total leakage amount. The current source differentiation module is used to distinguish currents from different sources based on the characteristics of the zero-sequence current information, in order to avoid protection actions caused by the superposition of non-faulty currents; and The protection action triggering module is used to trigger corresponding protection actions when a leakage risk is detected.

[0008] Compared with the prior art, this application has at least the following beneficial effects: This application reconstructs the true total leakage current of multiple power circuits by acquiring zero-sequence current information, electrical connection relationships, and current distribution ratios. By accurately calculating the true total leakage current, it is possible to accurately determine whether a leakage risk exists, thus avoiding the risk of workers being exposed to dangerous contact voltages for extended periods.

[0009] Furthermore, this application distinguishes currents from different sources based on the characteristics of zero-sequence current information to avoid protection actions caused by the superposition of non-faulty currents. Therefore, it can ensure that the protection system can act in a timely manner when a leakage actually occurs, while avoiding critical operation interruptions, economic losses and project delays caused by false actions.

[0010] In summary, this application can adapt to systemic electrical connection changes caused by temporary operations, and provides a dynamic and global risk assessment and hierarchical protection strategy, which significantly improves the safety, reliability and efficiency of power operations, and effectively solves the serious problem of protection systems failing or malfunctioning at critical moments in the prior art. Attached Figure Description

[0011] Figure 1 This application provides a flowchart illustrating a risk assessment-based method for hierarchical configuration of leakage current protection in electrical work.

[0012] Figure 2 This application provides a structural schematic diagram of a risk assessment-based graded configuration system for leakage current protection in power operations. Detailed Implementation

[0013] The technical solutions in this application will now be clearly and completely described in conjunction with the accompanying drawings.

[0014] This application proposes a risk-assessment-based method for hierarchical configuration of leakage current protection in power operations, such as... Figure 1 As shown, it includes the following steps: Obtain zero-sequence current information from multiple power circuits; Obtain the electrical connection relationships and current distribution ratios between multiple power circuits; Based on zero-sequence current information, electrical connection relationships, and current distribution ratios, the actual total leakage current of multiple power circuits is reconstructed. Based on the actual total leakage current, determine whether there is a risk of leakage current. Based on the characteristics of zero-sequence current information, currents from different sources are distinguished to avoid protection actions caused by the superposition of non-fault currents; and When a risk of leakage is detected, the corresponding protection action is triggered.

[0015] This application dynamically acquires the electrical connection relationships and current distribution ratios of power circuits, and reconstructs the true total leakage current by combining it with zero-sequence current information, thereby accurately determining leakage risk. Simultaneously, this application can distinguish currents from different sources, effectively avoiding protection malfunctions caused by the superposition of non-fault currents, and significantly improving the reliability and safety of leakage protection systems in complex power operation environments.

[0016] To better understand this application, some key terms involved need to be explained.

[0017] Zero-sequence current information refers to the vector sum of the three-phase currents in a power system, which theoretically has a value of zero under normal operating conditions. When a single-phase ground fault or leakage occurs, the zero-sequence current information will no longer be zero, and its magnitude reflects the severity of the leakage. A power circuit refers to the path of current flow in a power system, which can include power sources, conductors, loads, and various protection and control devices. Electrical connection relationships describe how different power circuits are interconnected, such as series, parallel, or star connections. Current distribution ratio refers to the distribution of the total current among the branches in parallel or branch paths. Actual total leakage current refers to the sum of the actual leakage current occurring in the entire power system after considering all branch paths and electrical connection relationships. Non-fault current refers to current not caused by insulation damage, such as common-mode current generated during normal equipment operation and induced current caused by electromagnetic interference. Protection action refers to the measures taken by the leakage protection device, such as cutting off the power supply and issuing an alarm, when a leakage risk is detected.

[0018] This application provides a risk assessment-based method for hierarchical configuration of leakage protection in power operations. Its core lies in accurately assessing and effectively responding to dynamically changing electrical connections and leakage conditions in complex power operation environments.

[0019] First, this method requires acquiring zero-sequence current information from multiple power circuits. This zero-sequence current information can be monitored in real time by installing zero-sequence current transformers in each power circuit. For example, a ring-shaped zero-sequence current transformer can be installed at the distribution box or the incoming line of critical equipment in each work area. When current flows through, the transformer will induce a voltage or current signal proportional to the zero-sequence current. These signals are then collected and digitized to form zero-sequence current information data.

[0020] Secondly, this method requires obtaining the electrical connection relationships and current distribution ratios between multiple power circuits. Specifically, this can be achieved by manually inspecting and consulting electrical drawings to obtain the initial electrical connection relationships, and then estimating the current distribution ratio based on the circuit's impedance characteristics. For example, before each operation begins, a professional electrician inspects the temporary wiring on site, records the connection method and conductor specifications used for each circuit, and then calculates the resistance based on the conductor length and cross-sectional area, thereby estimating the current distribution ratio.

[0021] Next, based on the zero-sequence current information, electrical connection relationships, and current distribution ratios, the actual total leakage current of multiple power circuits is reconstructed. Specifically, this can be achieved by simply adding the zero-sequence current information measured for each circuit and then correcting it according to a preset empirical coefficient. For example, if the zero-sequence current information of one circuit is known to be I1 and the zero-sequence current information of another circuit is known to be I2, and there is a parallel relationship between them, then I1 and I2 can be simply added together to obtain a preliminary estimate of the total leakage current.

[0022] Then, based on the actual total leakage current, it is determined whether there is a leakage risk. Specifically, this can be done by comparing the reconstructed actual total leakage current with a preset fixed threshold. For example, if the preset leakage risk threshold is 30mA, then if the reconstructed actual total leakage current exceeds 30mA, then a leakage risk is determined to exist.

[0023] Furthermore, based on the characteristics of zero-sequence current information, currents from different sources can be distinguished to avoid protection actions caused by the superposition of non-faulty currents. Specifically, this can be done through simple amplitude comparison. For example, if the detected zero-sequence current information has a small amplitude and a short duration, it may be considered a transient interference or a non-faulty current; if the amplitude is large and the duration is long, it may be considered a real leakage current.

[0024] Finally, when a leakage risk is detected, the corresponding protective action is triggered. Specifically, when a leakage risk is detected, a trip command is sent directly to the circuit breakers of all relevant circuits to cut off the power supply. For example, if the system determines that there is a leakage risk in a certain area, the main power supply to that area is immediately disconnected to prevent electric shock accidents.

[0025] This application reconstructs zero-sequence current information by dynamically acquiring and utilizing the electrical connection relationships and current distribution ratios of power circuits, thereby obtaining a more accurate total leakage current. Compared to traditional systems that rely solely on local zero-sequence current detection, this approach provides a more comprehensive assessment of the leakage status of the entire power network. For example, in a traditional system, if a leakage occurs in work area A, but due to temporary connections, the leakage current is diverted to work area B, the protection device in work area A may fail because the detected current is below a threshold. However, this application, by acquiring the electrical connection relationships and current distribution ratios between work areas A and B, can take the diverted current into account, reconstructing the true total leakage current, thereby enabling timely detection and handling of leakage risks in work area A.

[0026] Furthermore, this application introduces a mechanism to distinguish currents from different sources based on the characteristics of zero-sequence current information. This enables the system to effectively identify and eliminate interference from non-faulty currents (such as common-mode current generated during normal equipment operation, electromagnetic interference, etc.) on leakage protection judgment. For example, in a traditional system, the leakage current from the grinding wheel in work area A and the common-mode leakage current generated by the precision diagnostic equipment in work area B are superimposed at the leakage protection device in work area B, which may lead to malfunction of the protection device in work area B. By analyzing the characteristics of zero-sequence current information, this application can distinguish non-faulty currents such as common-mode current, avoiding their superposition with the actual leakage current and resulting in misjudgment, thereby improving the reliability of the protection system and the continuity of operation.

[0027] However, in actual power systems, due to the existence of complex grounding networks and parallel grounding paths that may not be perceived by the system, traditional acquisition methods may find it difficult to accurately quantify the shunt capacity of these paths, thus affecting the accuracy of reconstructing the true total leakage current.

[0028] In response, this application further proposes a more accurate method for obtaining the aforementioned electrical connection relationships and current distribution ratios, namely: The steps for obtaining the electrical connection relationships and current distribution ratios among the above-mentioned multiple power circuits include: Inject coded broadband probe current pulses into the grounding loops of multiple power circuits; Monitor the zero-sequence current response of multiple power circuits; The total injected probe current energy is compared with the sum of the probe current energies detected by all known loops to quantify the shunt capacity of parallel ground paths not perceived by the system. Based on the quantified shunt capacity, estimate the equivalent shunt impedance of the parallel connection ground path; Reconstruct the current shunting ratio between all known paths and between known paths and parallel ground paths.

[0029] Specifically, injecting coded broadband probe current pulses into the grounding loops of multiple power circuits involves applying a current signal with a specific frequency range and coding pattern to the grounding system of the power circuits using specialized injection equipment. The coding characteristics of this probe current pulse make it easily identifiable and distinguishable in complex electromagnetic environments, while its broadband nature facilitates comprehensive detection of loop responses at different frequencies. The aim is to actively detect and stimulate the response of the grounding loops for subsequent analysis.

[0030] Monitoring the zero-sequence current response of multiple power circuits can be understood as simultaneously injecting probe current pulses and using high-precision zero-sequence current sensors to collect and record the zero-sequence current in each power circuit in real time. These zero-sequence current responses contain information on the propagation and shunting of the probe current along different paths.

[0031] In practical applications, comparing the total injected probe current energy with the sum of probe current energies detected in all known circuits to quantify the shunt capacity of undetected parallel grounding paths involves comparing the energy of the total injected probe current pulse with the energy of the probe current responses detected in all known, monitored power circuits. If an energy difference exists, this difference is considered to be the probe current energy flowing through undetected parallel grounding paths, thereby quantifying the shunt capacity of these undetected paths. The aim is to reveal and quantify hidden grounding paths that are difficult to detect using traditional methods.

[0032] Furthermore, estimating the equivalent shunt impedance of the parallel ground paths based on the quantified shunt capacity refers to calculating the equivalent impedance of these parallel ground paths using circuit theory, based on the quantified shunt capacity and the voltage or current characteristics of the injected probe current. This equivalent impedance reflects the shunt capability of these paths for leakage current.

[0033] Therefore, reconstructing the current shunting ratio between all known paths and between known paths and parallel ground paths refers to comprehensively considering the impedance characteristics of known loops, the shunting capacity of the quantified parallel ground paths and their equivalent impedance, and establishing a complete current shunting model to accurately calculate the specific distribution ratio of current between all known paths and undetected parallel ground paths under any given leakage current condition.

[0034] This application achieves comprehensive capture of current propagation paths in complex grounding networks by actively injecting coded broadband probe current pulses into grounding loops and simultaneously monitoring the zero-sequence current response of each power loop. By comparing the total injected probe current energy with the sum of the energy detected by known loops, the shunt capacity of parallel grounding paths not perceived by the system can be accurately quantified, thus overcoming the shortcomings of traditional methods in identifying hidden paths. Based on this, the equivalent shunt impedance of these parallel grounding paths is estimated, allowing these previously "invisible" paths to be incorporated into the current shunt model. Finally, by reconstructing the current shunt ratio including all known paths and parallel grounding paths, a comprehensive and accurate acquisition of the electrical connection relationships and current distribution ratios between power loops is ensured, effectively avoiding errors in total leakage current reconstruction due to incomplete information.

[0035] Suppose an industrial park contains multiple electrical circuits, some of which may have complex parallel grounding paths formed by underground metal pipes, reinforced concrete structures, etc., and not all of these paths are detected by traditional monitoring systems. To accurately obtain the electrical connections and current distribution ratios of these circuits, the following steps can be taken: First, a probe current pulse with a specific code (e.g., pseudo-random sequence code) and a wide frequency range (e.g., 1 kHz to 100 kHz) is periodically injected into the grounding circuit of all power circuits in the park through a dedicated signal injection device.

[0036] Meanwhile, a highly sensitive zero-sequence current sensor is installed at a key monitoring point in each power circuit to synchronously monitor and record the zero-sequence current response after the probe current pulse is injected.

[0037] Next, the system calculates the energy of the total injected probe current pulses and compares it to the sum of the probe current response energies detected by all known loops (e.g., loops defined by the design drawings). If the injected energy is found to be greater than the sum of the energies detected by the known loops, the difference is quantified as the shunt capacity flowing through parallel ground paths not perceived by the system.

[0038] Based on the quantified shunt capacity and the voltage information of the injected probe current, the equivalent shunt impedance of these parallel ground paths can be estimated. For example, if the quantified shunt capacity is X amperes and the injected voltage is Y volts, the equivalent impedance can be estimated as Y / X ohms.

[0039] Finally, based on the impedance models of all known loops and the estimated equivalent shunt impedance of the parallel grounding paths, the system reconstructs a complete current shunt model, thereby accurately calculating the specific distribution ratio of current among all known paths and these undetected parallel grounding paths under any leakage condition. For example, when a leakage occurs in a loop, it can accurately predict how much leakage current will flow through the known grounding wire and how much will be shunt through concealed paths such as underground pipes.

[0040] In some implementations, it is proposed to distinguish currents from different sources based on the characteristics of zero-sequence current information in order to avoid protection actions caused by the superposition of non-faulty currents, i.e.: The steps described above, which distinguish currents from different sources based on the characteristics of zero-sequence current information to avoid protection actions caused by the superposition of non-fault currents, include: Spectral analysis is performed on zero-sequence current information to extract its frequency components and harmonic characteristics. The frequency components and harmonic characteristics are compared with the preset non-fault leakage current characteristic template; Transient waveform analysis is performed on zero-sequence current information to extract its transient characteristics; The transient characteristics are matched with a preset template of non-faulty leakage current transient characteristics; Multidimensional feature fusion analysis is performed based on frequency components, harmonic characteristics, comparison results, intensity of transient features, duration of transient features, matching results, propagation path of zero-sequence current information in different loops, and attenuation of zero-sequence current information in different loops. Based on the results of multidimensional feature fusion analysis, the sources of zero-sequence current are distinguished.

[0041] Specifically, spectral analysis of zero-sequence current information refers to converting the time-domain zero-sequence current signal to the frequency domain using signal processing techniques such as Fourier transform, thereby obtaining its frequency components and harmonic characteristics. The frequency components can be understood as the energy distribution of the zero-sequence current signal at different frequency points, while harmonic characteristics refer to the amplitude and phase of the integer multiples of the fundamental frequency in the signal. The purpose is to identify, from a frequency perspective, non-faulty currents that may be caused by normally operating equipment (such as frequency converters and switching power supplies) or harmonic pollution from the power grid.

[0042] Furthermore, comparing the frequency components and harmonic characteristics with a preset non-faulty leakage current characteristic template involves comparing the actually measured frequency and harmonic characteristics with pre-established characteristic patterns representing known non-faulty currents (e.g., leakage currents caused by capacitive loads, early insulation aging, electromagnetic interference, etc.). This comparison can be achieved using pattern recognition, machine learning algorithms, or simple threshold judgment, with the aim of initially screening out zero-sequence current components with non-faulty characteristics.

[0043] Furthermore, transient waveform analysis of zero-sequence current information refers to analyzing the rapid changes in the zero-sequence current signal over a short period of time and extracting its transient characteristics through methods such as wavelet transform, short-time Fourier transform, or empirical mode decomposition. Transient characteristics can include the rising edge, falling edge, pulse width, peak value, and oscillation frequency of the current. The aim is to capture short-duration, rapidly changing non-steady-state current events, which are often related to certain operations (such as switch closing or motor starting) or transient disturbances.

[0044] Matching transient features with a preset template of non-faulty leakage current transient features involves matching the extracted transient features with pre-stored feature patterns representing typical non-faulty transient currents (e.g., transient currents caused by surges, instantaneous load switching, arcing, etc.). Through matching, transient zero-sequence currents that correspond to known non-faulty events can be identified.

[0045] In practical applications, multi-dimensional feature fusion analysis is performed based on frequency components, harmonic characteristics, comparison results, the intensity and duration of transient features, matching results, the propagation path of zero-sequence current information in different loops, and the attenuation of zero-sequence current information in different loops. This can be understood as comprehensively utilizing feature information from different dimensions and analysis methods, and using advanced algorithms such as weighted summation, decision trees, neural networks, or Bayesian networks to comprehensively determine the source of zero-sequence current. For example, propagation paths and attenuation can provide information on the spatial distribution of current, which helps distinguish between local interference and systemic leakage. The aim is to improve the accuracy and robustness of the judgment and avoid the limitations of single-feature judgment.

[0046] Therefore, based on the results of multi-dimensional feature fusion analysis, the source of zero-sequence current can be distinguished, thereby accurately identifying the real leakage fault and eliminating the interference of non-fault current.

[0047] This application effectively solves the problem of protection malfunction caused by the superposition of non-faulty currents in traditional methods by conducting in-depth multi-dimensional analysis of zero-sequence current information. Specifically, firstly, through spectrum analysis, steady-state non-faulty currents caused by normally operating equipment or grid harmonics can be identified in the frequency domain, such as specific harmonic components generated by frequency converters or power frequency leakage current caused by capacitive loads. After comparing these frequency components and harmonic characteristics with preset non-faulty leakage current characteristic templates, some interference can be initially eliminated. Secondly, through transient waveform analysis, transient non-faulty currents with short duration and rapid changes can be captured, such as the instantaneous inrush current generated during switch operation or motor startup. After matching these transient characteristics with preset non-faulty leakage current transient characteristic templates, transient interference can be further identified and eliminated. Finally, and most importantly, through multi-dimensional feature fusion analysis, the analysis results in the frequency domain and time domain are combined with spatial information such as the intensity and duration of transient characteristics, as well as the propagation path and attenuation of zero-sequence current in different circuits, for comprehensive consideration. This fusion analysis leverages the complementarity between different characteristics. For example, a weak harmonic component alone may not be sufficient to trigger protection, but if it is accompanied by specific transient characteristics and its propagation path is consistent with a known interference source, it can be more accurately identified as a non-fault current. Through this comprehensive judgment mechanism, the system can more accurately identify actual leakage faults and avoid maloperation of protection caused by the superposition of non-fault currents.

[0048] However, in practical applications, relying solely on passive feature analysis of zero-sequence current information may not be sufficient to effectively identify and eliminate all complex and variable non-fault currents, especially those interference signals that are weak, intermittent, or similar to the characteristics of real leakage current, which may lead to misjudgment or missed judgment.

[0049] In response, this application further proposes a method for distinguishing currents from different sources, the steps of which include: For each power circuit, a weak test current with a specific frequency and waveform characteristics is periodically injected; Simultaneously monitor the zero-sequence current response after injecting the test current in all power circuits; By comparing the characteristics of the monitored zero-sequence current response with those of the injected test current, the response component caused by the test current is separated. The separated response components are removed from the total zero-sequence current information to obtain zero-sequence current information without the test current response; Residual signal analysis is performed on the zero-sequence current information after removing the test current response to identify whether there are weak, intermittent signals that match the preset fault characteristics. When a weak, intermittent signal matching the preset fault characteristics is identified in the residual signal, it is determined that a real leakage fault exists.

[0050] Specifically, a weak test current is periodically injected into each power circuit. The purpose is to establish a controllable reference signal through active probing, without affecting the normal operation of the power system. This test current typically has specific frequency and waveform characteristics; for example, it can be a high-frequency pulse signal or a sine wave of a specific frequency, to facilitate its differentiation from other zero-sequence current components in the system during subsequent monitoring. Periodic injection ensures continuous monitoring and dynamic evaluation of the circuit's condition.

[0051] The synchronous monitoring of the zero-sequence current response after the injection of test current in all power circuits refers to the precise recording of zero-sequence current changes in each circuit by deploying zero-sequence current sensors in each circuit simultaneously with or immediately after the injection of test current. Synchronous monitoring can capture the propagation path, attenuation, and mutual influence of the test current in different circuits, providing a comprehensive data foundation for subsequent signal separation.

[0052] In practical applications, the characteristics of the monitored zero-sequence current response are compared with those of the injected test current, and the response component caused by the test current is separated. Specifically, this involves using signal processing techniques, such as correlation analysis, digital filtering, or pattern recognition, to compare the monitored zero-sequence current signal with the known injected test current signal. Because the test current has unique characteristics, the specific response component caused by it can be effectively extracted from the complex total zero-sequence current.

[0053] Furthermore, the separated response components are removed from the total zero-sequence current information to obtain zero-sequence current information without the test current response. The purpose is to eliminate the interference of the actively injected test signal on the actual leakage current judgment, thereby obtaining a more "pure" zero-sequence current background, which mainly reflects the system's own operating status and potential leakage current.

[0054] Based on this, residual signal analysis is performed on the zero-sequence current information after removing the test current response to identify whether there are weak, intermittent signals consistent with preset fault characteristics. Residual signal analysis can employ various advanced signal processing algorithms, such as wavelet analysis, Fourier transform, or machine learning classifiers, to detect weak or intermittent leakage current signals with specific frequency, amplitude, or duration characteristics that might be overlooked by conventional methods. Preset fault characteristics are established based on historical data and expert experience to describe typical signal patterns of real leakage current faults.

[0055] Therefore, when a weak, intermittent signal matching the preset fault characteristics is identified in the residual signal, a real leakage fault can be determined. This judgment mechanism is based on the effective suppression of background noise and non-fault current, enabling even difficult-to-detect early or intermittent leakage to be accurately identified.

[0056] The proposed solution effectively decouples the system's own zero-sequence current information from the response component of the test current by actively injecting a weak test current with known characteristics and precisely monitoring and separating its response in the power circuit. This active detection mechanism enables the system to accurately isolate background noise and interference caused by non-fault factors (such as electromagnetic interference, load transients, etc.) and controllable signals generated by the test current itself from the complex zero-sequence current signal. By removing these known or controllable signal components, the resulting residual signal can more realistically and sensitively reflect whether there are unknown weak or intermittent leakage current signals in the power circuit that match the preset fault characteristics. This method avoids the protection malfunctions that may occur due to the superposition of non-fault currents in traditional passive monitoring, while improving the detection sensitivity and accuracy of real leakage current faults.

[0057] Assume there are multiple parallel-operating power circuits within a power operation area. To accurately distinguish the source of zero-sequence current, the system injects a weak sinusoidal test current with a frequency of 1kHz and an amplitude of 10mA into the grounding circuit of each power circuit, for example, every 5 minutes, via a dedicated injection device. Simultaneously, high-precision zero-sequence current sensors on all power circuits synchronously monitor their respective zero-sequence current information at a sampling rate of 10kHz. The data processing unit then receives this monitoring data. Using a digital bandpass filter with a center frequency of 1kHz, the total zero-sequence current information of each circuit is filtered to extract the response component corresponding to the frequency of the injected test current. For example, if the total zero-sequence current monitored by circuit A is I_total_A, the response component I_test_A caused by the test current can be separated through filtering and correlation analysis. Then, I_test_A is subtracted from I_total_A to obtain the zero-sequence current information I_residual_A, which does not contain the test current response. Wavelet transform analysis is performed on I_residual_A to identify transient pulse signals with a duration of less than 100ms, a frequency between 50Hz and 150Hz, and an amplitude greater than 5mA. These signals are pre-defined as typical characteristics of intermittent leakage faults. If a signal matching these characteristics is detected in I_residual_A, the system determines that a real leakage fault exists in loop A and immediately triggers corresponding graded protection actions, such as issuing an early warning first, and further isolating the loop if the signal persists or intensifies. Through this active detection and precise separation method, even in the presence of complex electromagnetic interference or load fluctuations, false judgments can be effectively avoided, and potential leakage risks can be detected in a timely manner.

[0058] However, in actual power operation environments, non-fault currents (such as electromagnetic interference, harmonic currents, etc.) often have complex time-varying and dynamic characteristics. Their characteristics may overlap or be ambiguous with those of the actual leakage current, making it difficult to accurately identify them based on a single characteristic, which may lead to malfunctions or failures to operate of protection devices.

[0059] To address this, this application proposes a more refined and robust method for distinguishing zero-sequence current sources. Through joint time-frequency analysis and multi-dimensional information fusion, it effectively improves the identification accuracy of non-faulty currents, namely: The steps described above for distinguishing currents from different sources to avoid protection actions caused by the superposition of non-fault currents include: Time-frequency joint analysis of zero-sequence current information is performed to generate the time-frequency distribution of zero-sequence current information; The energy distribution, instantaneous frequency change rate of zero-sequence current information, and the time trajectory of harmonic components of zero-sequence current information are extracted from the time-frequency distribution. The energy distribution, instantaneous frequency change rate, and harmonic component change trajectory over time are compared with the preset electromagnetic interference characteristic time-frequency template to adapt to the distortion and drift of the characteristics on the time axis. By combining the load change information of the power circuit with the operating status of the equipment in the power circuit, the results are compared and corrected; Based on the corrected comparison results, the source of zero-sequence current is identified.

[0060] Specifically, time-frequency joint analysis of zero-sequence current information refers to using time-frequency analysis techniques such as Short-Time Fourier Transform (STFT), Wavelet Transform (WT), or Hilbert-Huang Transform (HHT) to transform the original zero-sequence current time-domain signal into the time-frequency domain, generating a time-frequency distribution of the zero-sequence current information. This time-frequency distribution can simultaneously show the energy distribution of the zero-sequence current signal at different times and frequencies, thus revealing its dynamic spectral characteristics. Extracting the energy distribution, instantaneous frequency change rate, and harmonic component variation trajectory of the zero-sequence current information from the time-frequency distribution can be understood as feature extraction from the generated time-frequency distribution map. The energy distribution reflects the signal strength at a specific time-frequency point; the instantaneous frequency change rate describes how quickly the signal frequency changes over time, which is important for identifying transient interference or fault signals; the harmonic component variation trajectory can track the intensity and frequency drift of specific harmonic components at different times, which is crucial for distinguishing between harmonic currents caused by nonlinear loads and actual leakage currents.

[0061] In practical applications, the extracted energy distribution, instantaneous frequency change rate, and harmonic component variation trajectories over time are compared with preset electromagnetic interference characteristic time-frequency templates. The aim is to identify and eliminate non-faulty currents. These templates are built upon extensive historical data and experience, containing unique fingerprints of various typical electromagnetic interferences (such as switching operations, motor starts, and inverter operation) in the time-frequency domain. During the comparison process, algorithms such as Dynamic Time Warping (DTW) are employed to adapt to distortions and drifts of features on the time axis, ensuring accurate identification even if the characteristics of non-faulty currents are distorted to some extent. Furthermore, the comparison results are corrected by combining load change information and equipment operating status information of the power circuit. For example, when the load of a power circuit suddenly increases or a large piece of equipment starts, transient zero-sequence currents may be generated; these currents are not leakage faults. By acquiring this load and equipment operating status information in real time, the time-frequency comparison results can be re-verified, eliminating non-faulty currents caused by normal operation or equipment operation, thereby improving the accuracy of the judgment. Therefore, based on the corrected comparison results, the source of zero-sequence current can be accurately distinguished, and it can be determined whether it originates from a real leakage fault or is caused by non-fault factors such as electromagnetic interference and harmonic current.

[0062] This application's solution, by introducing joint time-frequency analysis, extends the analysis of zero-sequence current signals from a single time or frequency domain to a two-dimensional time-frequency space, thereby capturing the dynamic characteristics of the signal more comprehensively. Traditional single-domain analysis methods struggle to effectively distinguish between non-fault currents and true leakage currents that have similar time-domain waveforms but different frequency characteristics, or similar frequency-domain components but different temporal behaviors. For example, some transient electromagnetic interferences may overlap with the harmonic components of leakage current in the frequency domain but exhibit an extremely short duration in the time domain; while some steady-state harmonic currents may persist in the time domain but have specific, non-leakage-related frequency distributions in the frequency domain. Through joint time-frequency analysis, the time and frequency characteristics of these signals can be observed simultaneously, allowing currents from different sources to exhibit unique "fingerprints" in their time-frequency distributions, thus improving the accuracy of differentiation. Furthermore, comparing the extracted time-frequency features with a preset electromagnetic interference feature time-frequency template and incorporating adjustments based on load changes and equipment operating conditions further enhances the system's robustness. Electromagnetic interference (EMI) often exhibits time-varying and uncertain characteristics. By comparing templates and adapting to their distortion and drift along the time axis, these complex non-fault signals can be effectively identified. Simultaneously, corrections are made based on actual operating condition information, avoiding misjudgments caused by transient currents during normal operation or equipment switching. This ensures that a leakage risk is only identified when the characteristics of the zero-sequence current highly match the actual leakage fault and all known non-fault factors are excluded, thus effectively preventing protection actions caused by the superposition of non-fault currents.

[0063] The following is a specific example to illustrate this.

[0064] Suppose that in a power operation area, a zero-sequence current sensor in a power circuit 100 detects continuous zero-sequence current information. First, this zero-sequence current information is sent to a time-frequency analysis unit, where a time-frequency distribution map is generated through wavelet transform. From this time-frequency distribution map, the system extracts the energy distribution, instantaneous frequency change rate, and the trajectory of major harmonic components (such as the 3rd and 5th harmonics) of the zero-sequence current information at different time points over time. Subsequently, these extracted time-frequency features are sent to a feature comparison module for comparison with a pre-defined electromagnetic interference feature time-frequency template library. For example, the template library may contain time-frequency templates for specific frequency sweep signals generated when a frequency converter starts up, or time-frequency templates for broadband noise generated when an electric arc furnace operates. The comparison algorithm (such as dynamic time warping) evaluates the similarity between the current zero-sequence current's time-frequency features and these templates, even if the signal is stretched or compressed on the time axis. Simultaneously, the system obtains current load change information and equipment operating status from the monitoring system of power circuit 100. If the comparison results show that the time-frequency characteristics of the zero-sequence current highly match a certain electromagnetic interference template, and simultaneously, equipment related to that electromagnetic interference source is detected to be operating (e.g., a frequency converter is starting up), then the comparison results will be corrected, determining that the zero-sequence current mainly originates from non-faulty electromagnetic interference. Conversely, if the comparison results have a low degree of matching with the electromagnetic interference template, or although they match, there is no corresponding equipment operating status to support it, the system will further analyze whether it conforms to the time-frequency characteristics of a real leakage fault. Through this multi-dimensional, dynamic analysis and correction mechanism, real leakage current and non-faulty current can be accurately distinguished, thereby avoiding unnecessary protection actions.

[0065] However, in actual power operation environments, non-fault currents (such as currents caused by equipment operation, electromagnetic interference, or system imbalance) often have complex spatial distribution and temporal dynamic variation characteristics. If only a single loop or simple characteristics are relied upon for differentiation, it may be difficult to accurately identify these non-fault currents with spatial differences and temporal dynamic coupling characteristics, which may lead to false triggering or missed triggering of protection actions.

[0066] To address this, this application further proposes a more refined method for distinguishing current sources. By performing comprehensive spatiotemporal correlation analysis on zero-sequence current information, it can more accurately identify and differentiate currents from different sources, especially non-fault currents with complex spatiotemporal characteristics, namely: The steps described above for distinguishing currents from different sources to avoid protection actions caused by the superposition of non-fault currents include: Spatial correlation analysis was performed on zero-sequence current information obtained from power circuits in different work areas to identify the propagation characteristics and mutual influence relationships of zero-sequence current information between different work areas. Based on the spatial correlation analysis results, time series analysis was performed on the zero-sequence current information to extract the dynamic characteristics of the zero-sequence current information changing with time and the correlation at different time points. By combining the results of spatial correlation analysis and temporal analysis, a spatiotemporal correlation diagram of zero-sequence current information is constructed; Extracting the spatiotemporal differences and spatiotemporal coupling characteristics of zero-sequence current information from the spatiotemporal correlation graph; The spatiotemporal difference characteristics and spatiotemporal coupling characteristics are compared with the preset spatiotemporal characteristic template of non-faulty current to identify non-faulty currents with spatial differences and dynamic temporal coupling characteristics. Based on the comparison results, the source of the zero-sequence current can be identified.

[0067] Specifically, spatial correlation analysis refers to the comprehensive analysis of zero-sequence current information obtained from power circuits in multiple geographically dispersed or electrically interconnected work areas. The aim is to reveal the spatial distribution patterns, mutual influence paths, and propagation characteristics of these currents. For example, spatial dependencies can be quantified by calculating the cross-correlation coefficients and coherence between zero-sequence currents in different circuits, or by constructing spatial connectivity models using graph theory methods. The goal is to understand how zero-sequence currents diffuse or concentrate in different regions, thus providing a spatial basis for subsequent differentiation.

[0068] Time series analysis can be understood as an in-depth investigation into the changes in zero-sequence current information over time, based on the results of spatial correlation analysis. This includes extracting the dynamic characteristics of the zero-sequence current, such as the trends in its amplitude, frequency, and phase over time, as well as its autocorrelation or cross-correlation at different time points. For example, methods such as Fourier transform, wavelet analysis, or empirical mode decomposition can be used to analyze the periodicity, transient changes, or trends of the zero-sequence current. Its purpose is to capture the temporal evolution of the zero-sequence current, especially the specific time patterns that non-fault currents may exhibit.

[0069] In practical applications, constructing a spatiotemporal correlation graph for zero-sequence current information by combining spatial correlation analysis and temporal analysis results refers to integrating measurement data of zero-sequence current at different spatial locations and time points into a unified representation framework. This spatiotemporal correlation graph can be a multidimensional matrix, a graph structure, or a tensor, where each element or node represents the zero-sequence current information at a specific spatiotemporal point, and connections represent their spatiotemporal correlations. Its purpose is to provide a comprehensive view to understand the overall behavior of zero-sequence current from both macroscopic and microscopic perspectives.

[0070] Furthermore, extracting the spatiotemporal differences and spatiotemporal coupling characteristics of zero-sequence current information from the spatiotemporal correlation graph refers to identifying features with unique spatiotemporal patterns from the constructed spatiotemporal correlation graph using specific algorithms and models. Spatiotemporal differences may manifest as abnormal fluctuations in zero-sequence current within a specific region or time period, or significant differences in zero-sequence current between different regions. Spatiotemporal coupling characteristics refer to the mutual influence and coordinated changes of zero-sequence current in spatial and temporal dimensions, such as how a current change in one region triggers a current response in another region, and the time delay of this response. The aim is to quantify the complex spatiotemporal behavior of zero-sequence current and provide key criteria for accurately distinguishing current sources.

[0071] Finally, the spatiotemporal differences and spatiotemporal coupling characteristics are compared with preset spatiotemporal feature templates for non-faulty currents to identify non-faulty currents with spatial differences and dynamic temporal coupling characteristics. This involves establishing a library of typical feature patterns of known non-faulty currents (such as harmonics, surges, and equipment starting currents) under different spatiotemporal conditions. Using techniques such as pattern recognition, machine learning, or deep learning, spatiotemporal features extracted from actual measurement data are matched with these templates. The aim is to accurately separate currents with complex spatiotemporal characteristics caused by non-faulty factors from potential real leakage currents, thereby improving the accuracy of the protection system.

[0072] This application overcomes the limitations of traditional methods in handling complex, distributed non-faulty currents by introducing spatial correlation analysis and temporal analysis, and further constructing a spatiotemporal correlation graph of zero-sequence current information. Traditional methods may only focus on the current characteristics of a single loop or simple time-domain / frequency-domain analysis, making it difficult to capture the propagation characteristics, mutual influence relationships, and dynamic characteristics of non-faulty currents across different operating areas. Because non-faulty currents often have specific spatiotemporal propagation paths and temporal evolution patterns—for example, some electromagnetic interferences may exhibit high intensity in specific areas and fluctuate periodically or randomly over time—while actual leakage currents may persist continuously at specific locations and possess different spatiotemporal characteristics.

[0073] This application utilizes spatial correlation analysis of zero-sequence current information to identify the propagation paths and mutual influences of current across different regions, thereby understanding the "geographical" distribution and diffusion patterns of current. Based on this, temporal analysis captures the dynamic characteristics of current changes over time and the correlations at different points in time, revealing the "temporal" evolution patterns of the current. Combining these two approaches to construct a spatiotemporal correlation diagram allows for a comprehensive examination of the overall behavior of zero-sequence current within a unified framework. Consequently, the spatiotemporal differences and coupling characteristics extracted from the spatiotemporal correlation diagram can more comprehensively and precisely characterize the complex spatiotemporal fingerprint of zero-sequence current. For example, a transient device startup current may momentarily exhibit a high amplitude in a specific region and rapidly decay, its spatiotemporal characteristics significantly different from those of a continuous leakage current. By comparing these refined spatiotemporal characteristics with a preset spatiotemporal characteristic template for non-faulty currents, the system can more accurately identify non-faulty currents with spatial differences and dynamic temporal coupling characteristics, thus avoiding protection malfunctions caused by their superposition.

[0074] Therefore, this application can significantly reduce protection malfunctions caused by the superposition of non-faulty currents, improving the reliability and availability of the protection system. Simultaneously, by more accurately identifying non-faulty currents, the system can focus more on detecting actual leakage faults, thereby enhancing its early warning capability for potential safety risks. This refined spatiotemporal analysis method provides solid technical support for achieving high-precision, high-reliability leakage protection in complex and ever-changing power operating environments.

[0075] Suppose a large power work site comprises multiple interconnected work areas, each with its own independent power circuit and different load devices. At a certain moment, the zero-sequence current information acquisition module detects zero-sequence current in multiple circuits. If judgment is made solely based on the amplitude or frequency of the zero-sequence current, it may be difficult to distinguish whether these currents are genuine leakage currents, transient startup of large equipment within a certain area, or electromagnetic interference across areas.

[0076] At this point, the solution proposed in this application will demonstrate its advantages. First, spatial correlation analysis is performed on the zero-sequence current information obtained from power circuits in different operating areas. For example, by analyzing the phase difference and amplitude ratio of the zero-sequence current in different circuits, it is found that the zero-sequence current in region A has a significant propagation delay and attenuation compared to the zero-sequence current in region B, while exhibiting high synchronization with the zero-sequence current in region C. This indicates that there may be some electrical connection or interference path between regions A and B, while regions A and C may share a common interference source.

[0077] Next, based on the spatial correlation analysis results, a time-series analysis was performed on the zero-sequence current information. For example, it was found that the synchronous zero-sequence currents in regions A and C exhibited periodic high-frequency pulse characteristics in time, and their duration coincided with the start-up cycle of a large welding equipment.

[0078] Subsequently, combining the results of spatial correlation analysis and temporal analysis, a spatiotemporal correlation diagram of zero-sequence current information was constructed. This diagram clearly shows how high-frequency pulse current propagates from region C to region A within a specific time period, exhibiting a specific spatial distribution and temporal evolution pattern.

[0079] From the spatiotemporal correlation diagram, the spatiotemporal difference characteristics of the high-frequency pulse current (e.g., highest intensity in region C, gradually attenuating outwards) and spatiotemporal coupling characteristics (e.g., highly synchronized with the welding equipment startup time and along a specific propagation path) are extracted. These characteristics are then compared with a preset spatiotemporal characteristic template of non-faulty current for "welding equipment startup current". If the comparison results show a high degree of match, the system can accurately determine that the zero-sequence current is not a real leakage current, but a non-faulty current caused by the startup of the welding equipment.

[0080] In this way, the system avoids triggering unnecessary protection actions due to misjudgment, ensuring the continuity and safety of power operations, while focusing attention on identifying the real risk of leakage.

[0081] Specifically, the steps for extracting the spatiotemporal difference characteristics and spatiotemporal coupling properties of zero-sequence current information from the spatiotemporal correlation graph include: For the spatiotemporal correlation graph, the spatiotemporal correlation graph is decomposed at different spatiotemporal scales; To obtain the energy distribution, frequency components, and propagation path characteristics of zero-sequence current at different scales; Spatiotemporal convolution operations are used to aggregate energy distribution, frequency components, and propagation path features; Extract the spatiotemporal difference characteristics and spatiotemporal coupling characteristics of zero-sequence current information.

[0082] The spatiotemporal correlation diagram is constructed by performing spatial correlation and temporal analysis on zero-sequence current information obtained from power circuits in different operating areas. It intuitively reflects the interrelationships of zero-sequence current information at different spatial locations and time points. Decomposing this spatiotemporal correlation diagram at different spatiotemporal scales involves using multi-scale analysis methods, such as wavelet decomposition and multi-resolution analysis, to break down the complex diagram into components within different time windows and spatial ranges. This decomposition helps reveal the behavioral patterns of zero-sequence current information at both macroscopic and microscopic levels, for example, identifying short-term transient events or long-term persistent changes.

[0083] Furthermore, at different scales, the energy distribution, frequency components, and propagation path characteristics of zero-sequence current information can be obtained. Specifically, energy distribution refers to the intensity or power distribution of the zero-sequence current signal at a specific spatiotemporal scale, reflecting the activity level of leakage events; frequency components refer to the spectral characteristics of the zero-sequence current signal at different spatiotemporal scales, which can reveal the periodicity and harmonic content of the signal, helping to distinguish currents from different sources; propagation path characteristics refer to the propagation trajectory and attenuation characteristics of the zero-sequence current signal from its source to the monitoring point in the power circuit network, which is crucial for locating leakage sources and understanding the behavior of current in complex networks.

[0084] Building upon this foundation, spatiotemporal convolution operations are used to aggregate energy distribution, frequency components, and propagation path features. Spatiotemporal convolution is a mathematical operation that fuses information across spatial and temporal dimensions. It involves sliding a learnable convolution kernel across a spatiotemporal correlation graph to perform a weighted summation of energy distribution, frequency components, and propagation path features within a local spatiotemporal region, thereby extracting higher-level, representative spatiotemporal features. The aggregation process aims to integrate information obtained from different scales and feature dimensions to form a unified and more generalized feature representation. Ultimately, through this aggregation process, the spatiotemporal variability and spatiotemporal coupling characteristics of zero-sequence current information can be extracted. Spatiotemporal variability refers to the unique patterns or changes exhibited by zero-sequence current information at different spatial locations or time points, while spatiotemporal coupling characteristics refer to the interrelationships and influences between zero-sequence current information at different spatial locations or time points.

[0085] This application decomposes the spatiotemporal correlation graph at different spatiotemporal scales, enabling the capture of the complex behavior of zero-sequence current information at both macroscopic and microscopic levels. Specifically, multi-scale decomposition allows the system to identify zero-sequence current patterns within different time windows and spatial ranges, such as transient leakage currents occurring over short periods or weak leakage currents persisting over long periods. By acquiring the energy distribution, frequency components, and propagation path characteristics at these different scales, the intrinsic properties of zero-sequence current can be comprehensively characterized. Furthermore, by aggregating these multi-scale, multi-dimensional features using spatiotemporal convolution operations, spatial and temporal information can be effectively fused to capture the local correlation and global dependence of zero-sequence current in the spatiotemporal dimension. This aggregation process helps filter out noise, enhance key features, and extract spatiotemporal differences and spatiotemporal coupling characteristics that better reflect the true nature of leakage currents or non-faulty currents. For example, by weighting the convolution kernels, abnormal energy concentrations or frequency changes in specific spatiotemporal regions can be highlighted, thereby more accurately identifying potential leakage risks and distinguishing them from background noise or non-faulty interference.

[0086] However, in practical applications, zero-sequence current information in power systems may contain complex and dynamically changing characteristics, such as transient high energy concentrations, drastic frequency changes, or nonlinear correlations between different scales. If a spatiotemporal convolution kernel with fixed size and weight is used, it may be difficult to effectively capture these delicate and variable characteristics, thus affecting the accurate differentiation of the source of zero-sequence current.

[0087] In this regard, this application further proposes the following steps for aggregating energy distribution, frequency components, and propagation path features using spatiotemporal convolution operations: For the spatiotemporal correlation graph, multi-scale decomposition is performed to obtain the energy distribution, frequency components and propagation path characteristics of zero-sequence current at different scales; Based on the energy distribution, frequency components, and propagation path characteristics of zero-sequence current information at different scales, the size and weight of the spatiotemporal convolution kernel are dynamically adjusted. When the zero-sequence current information in a specific spatiotemporal region is detected to exhibit high-intensity energy concentration or drastic frequency changes, the spatiotemporal convolution kernel size of that region is increased, and its weights are improved. When a nonlinear correlation is detected between zero-sequence current information at different scales, the weight allocation of the spatiotemporal convolution kernel is adjusted to strengthen the aggregation of nonlinear correlation features and weaken the aggregation of linear correlation features.

[0088] Specifically, based on multi-scale decomposition of the spatiotemporal correlation graph and the acquisition of energy distribution, frequency components, and propagation path characteristics of zero-sequence current information at different scales, this application introduces a mechanism for dynamically adjusting the size and weights of the spatiotemporal convolution kernel. "Dynamically adjusting" refers to adaptively changing the parameters of the convolution kernel according to the real-time characteristics of the zero-sequence current information to better match and extract features from the current data. When the system detects high-intensity energy concentration or drastic frequency changes in the zero-sequence current information in a specific spatiotemporal region, this usually indicates the presence of significant transient events or local anomalies in that region. To capture this key information more precisely, the size of the spatiotemporal convolution kernel is increased to cover a wider spatiotemporal range, while its weights are increased to enhance the response and aggregation capabilities to these high-intensity or drastic changing features. Furthermore, when nonlinear correlations are detected between zero-sequence current information at different scales, this means that a simple linear model is insufficient to describe these complex correlations. In this case, the weight allocation of the spatiotemporal convolution kernel is adjusted to strengthen the aggregation of nonlinear correlation features while weakening the aggregation of linear correlation features. The aim is to enable convolution operations to more effectively identify and extract nonlinear patterns hidden in complex data, which are crucial for distinguishing zero-sequence currents from different sources.

[0089] This application effectively addresses the limitations of fixed convolutional kernels in processing complex and dynamic zero-sequence current information by introducing a dynamic adjustment mechanism for spatiotemporal convolutional kernels. Specifically, when zero-sequence current information exhibits high-intensity energy concentration or drastic frequency changes in a specific spatiotemporal region, this is often a significant characteristic of faults or strong interference. By increasing the size of the spatiotemporal convolutional kernel in this region and enhancing its weights, the system can analyze these key regions more focusedly and deeply, capturing fine transient features that traditional fixed convolutional kernels might overlook, thereby improving sensitivity to abnormal events. Simultaneously, addressing the nonlinear correlations between different scales, the system strengthens the aggregation of nonlinear correlation features by adjusting the weight allocation of the convolutional kernels. This enables the system to identify and utilize deeper and more complex spatiotemporal patterns, which are crucial for distinguishing zero-sequence currents caused by different factors such as equipment aging, electromagnetic interference, or actual leakage. This adaptive aggregation method makes the feature extraction process more targeted and robust.

[0090] Through the above technical solution, this application can adaptively adjust the parameters of the spatiotemporal convolution kernel according to the real-time dynamic characteristics of zero-sequence current information, thereby significantly improving the ability to capture complex and dynamic zero-sequence current characteristics. Especially when facing complex scenarios such as high-intensity energy concentration, drastic frequency changes, or nonlinear correlations, this solution can more accurately extract key spatiotemporal difference features and spatiotemporal coupling characteristics, effectively avoiding the loss or misjudgment of feature information that may be caused by a fixed convolution kernel. Therefore, this application demonstrates higher accuracy and robustness in distinguishing zero-sequence currents from different sources, providing a more reliable data foundation for the risk assessment-based hierarchical configuration of leakage current protection in power operations.

[0091] Suppose that in a power circuit, the sudden start-up of a large inductive load causes a high-intensity transient spike in the zero-sequence current information within a short period of time, accompanied by significant frequency component changes. Traditional fixed spatiotemporal convolution kernels, due to their preset size and weights, may be unable to fully capture the fine structure of this transient spike and its rapid spatiotemporal evolution. According to the solution in this application, when the system detects this high-intensity energy concentration and drastically changing frequency zero-sequence current information, it immediately and dynamically increases the size of the spatiotemporal convolution kernel for that specific spatiotemporal region and increases its weights. For example, the receptive field of the convolution kernel can be expanded from a conventional 3x3 spatiotemporal unit to a 5x5 or 7x7 spatiotemporal unit, and its weight coefficient can be increased from 0.5 to 0.8. This allows the convolution operation to capture the waveform details, duration, propagation path, and attenuation of the transient spike in adjacent circuits more precisely. In this way, the system can more accurately identify that the zero-sequence current is a non-fault current caused by load start-up, rather than a true leakage fault, thereby avoiding unnecessary protection actions. For example, when the system analyzes zero-sequence current information, it may discover complex nonlinear relationships between the zero-sequence current responses of different power circuits at different time scales. This could indicate some form of distributed electromagnetic interference or multi-point grounding fault. In this case, the proposed solution adjusts the weight allocation of the spatiotemporal convolution kernel, for example, by introducing a nonlinear activation function or adjusting the weight matrix, to enhance the aggregation of these nonlinear correlation features. For instance, the weights of higher-order statistics or cross-correlation nonlinear terms can be increased, while the weights of simple linear correlation terms can be decreased. This adjustment enables the system to identify these nonlinear patterns, thereby more accurately distinguishing between complex electromagnetic interference and genuine leakage faults, and improving the intelligence level of leakage protection.

[0092] Specifically, when zero-sequence current information in a specific spatiotemporal region is detected to exhibit high-intensity energy concentration or drastic frequency changes, the steps of increasing the spatiotemporal convolution kernel size and improving its weights in that region may include the following: For each power circuit, the energy intensity and frequency change rate of zero-sequence current information are continuously monitored, and the timestamps and durations of its instantaneous changes are recorded. When high-intensity energy concentration or drastic frequency changes are detected in a specific spatiotemporal region, a high-speed feature sampling module is activated to collect zero-sequence current information in that region at a high sampling rate, obtaining a fine transient waveform. Based on the fine transient waveform, its transient energy distribution, instantaneous frequency change trajectory, and waveform distortion degree are calculated. According to the transient energy distribution, instantaneous frequency change trajectory, and waveform distortion degree, the dynamic size and weight coefficients of the spatiotemporal convolution kernel are calculated in real time. The dynamic size and weight coefficients are positively correlated with the complexity and duration of the transient features. In the spatiotemporal convolution operation, the spatiotemporal convolution kernel adjusted with dynamic size and weight coefficients is applied to aggregate the zero-sequence current information in a specific spatiotemporal region.

[0093] Continuous monitoring refers to the uninterrupted real-time data acquisition and analysis of the zero-sequence current in each power circuit. Energy intensity can be understood as the effective value or instantaneous power of the zero-sequence current, while the rate of frequency change reflects the stability of the current waveform. Recording the timestamps and durations of instantaneous changes aims to accurately track the occurrence and duration of abnormal events.

[0094] Furthermore, high-intensity energy concentration refers to a significant increase in the energy of the zero-sequence current within a short period of time, which may indicate a sudden fault or strong electromagnetic interference. Drastic frequency changes refer to a rapid and significant shift in the frequency components of the zero-sequence current. The high-speed feature sampling module is configured to acquire zero-sequence current information in this region at a rate far exceeding the conventional sampling rate when such anomalies are detected, thereby capturing fine transient waveform details that may be missed by traditional sampling methods.

[0095] Specifically, fine transient waveforms contain a wealth of fault or interference characteristics. Transient energy distribution refers to the distribution of energy at different frequencies or points in time over an extremely short timescale. Instantaneous frequency change trajectories describe the dynamic process of frequency change over time. Waveform distortion quantifies the degree to which the waveform deviates from a standard sine wave, for example, by measuring harmonic content or nonlinear distortion.

[0096] Therefore, the dynamic size of the spatiotemporal convolution kernel refers to the coverage of the kernel in both spatial and temporal dimensions, while the weight coefficients determine the importance of different data points in the convolution operation. Real-time computation means that these parameters can be adjusted instantly based on the latest transient feature analysis results. As a preferred implementation, the higher the complexity of the transient features (e.g., containing multiple frequency components, rapidly changing waveforms) and the longer their duration, the larger the convolution kernel size is required to cover a wider spatiotemporal range, and higher weights are assigned to emphasize their importance.

[0097] This application continuously monitors the zero-sequence current information of each power circuit and immediately activates a high-speed feature sampling module to acquire fine transient waveforms upon detecting high-intensity energy concentration or drastic frequency changes. By deeply analyzing these fine transient waveforms, calculating their transient energy distribution, instantaneous frequency change trajectory, and waveform distortion degree, the nature of abnormal events can be identified more accurately. Based on these precise transient features, the system can dynamically adjust the size and weights of the spatiotemporal convolution kernel in real time, making them positively correlated with the complexity and duration of the transient features. This adaptive adjustment mechanism ensures that the spatiotemporal convolution operation can more accurately focus on abnormal regions and optimize the granularity and focus of its analysis according to the specific attributes of the anomaly (such as intensity, duration, and complexity), thereby achieving effective aggregation and feature extraction of zero-sequence current information.

[0098] Through the above technical solution, the system can achieve more precise and real-time perception and response to abnormal zero-sequence current. This method of dynamically adjusting the spatiotemporal convolution kernel size and weights significantly improves the ability to distinguish between real leakage faults and non-faulty current disturbances in complex power environments. Especially when dealing with transient, intermittent, or complex waveform anomalies, it can effectively prevent power system instability caused by untimely or malfunctioning protection actions. Therefore, this solution improves the accuracy and reliability of leakage protection, reduces the false alarm rate, and enhances the system's adaptability to sudden abnormal events.

[0099] Based on the same inventive concept, this application also proposes a risk-assessment-based graded configuration system for leakage protection in electrical operations, such as... Figure 2 As shown, the system includes: Zero-sequence current information acquisition module 1 is used to acquire zero-sequence current information of multiple power circuits; Electrical connection relationship acquisition module 2 is used to acquire the electrical connection relationship and current distribution ratio between multiple power circuits; The leakage current total reconstruction module 3 is used to reconstruct the actual leakage current total of multiple power circuits based on zero-sequence current information, electrical connection relationship and current distribution ratio; The leakage risk assessment module 4 is used to determine whether there is a leakage risk based on the actual total leakage amount. The current source differentiation module 5 is used to differentiate currents from different sources based on the characteristics of zero-sequence current information, in order to avoid protection actions caused by the superposition of non-faulty currents; and The protection action triggering module 6 is used to trigger the corresponding protection action when it is determined that there is a risk of leakage.

[0100] This application dynamically acquires the electrical connection relationships and current distribution ratios of power circuits, and reconstructs the true total leakage current by combining it with zero-sequence current information, thereby accurately determining leakage risk. Simultaneously, the system can distinguish between currents from different sources, effectively avoiding protection malfunctions caused by the superposition of non-faulty currents, significantly improving the reliability and safety of leakage protection systems in complex power operation environments.

[0101] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A risk-assessment-based method for hierarchical configuration of leakage current protection in electrical work, characterized in that, include: Obtain zero-sequence current information from multiple power circuits; Obtain the electrical connection relationships and current distribution ratios among the multiple power circuits; Based on the zero-sequence current information, the electrical connection relationship, and the current distribution ratio, the actual total leakage current of the multiple power circuits is reconstructed. Based on the actual total leakage current, determine whether there is a risk of leakage current. Based on the characteristics of the zero-sequence current information, currents from different sources are distinguished to avoid protection actions caused by the superposition of non-faulty currents. as well as When a risk of leakage is detected, the corresponding protection action is triggered.

2. The method for hierarchical configuration of leakage protection for power operations based on risk assessment as described in claim 1, characterized in that, The step of obtaining the electrical connection relationship and current distribution ratio among the plurality of power circuits includes: Inject coded broadband probe current pulses into the grounding loops of the plurality of power circuits; Monitor the zero-sequence current response of the multiple power circuits; The total injected probe current energy is compared with the sum of the probe current energies detected by all known loops to quantify the shunt capacity of parallel ground paths not perceived by the system. Based on the quantified shunt capacity, estimate the equivalent shunt impedance of the parallel grounding path; Reconstruct the current shunting ratio between all known paths and between the known paths and the parallel grounding path.

3. The method for hierarchical configuration of leakage current protection for power operations based on risk assessment as described in claim 1, characterized in that, The step of distinguishing currents from different sources based on the characteristics of the zero-sequence current information to avoid protection actions caused by the superposition of non-fault currents includes: Spectral analysis is performed on the zero-sequence current information to extract the frequency components and harmonic characteristics of the zero-sequence current information; The frequency components and harmonic characteristics are compared with a preset non-faulty leakage current characteristic template; Transient waveform analysis is performed on the zero-sequence current information to extract its transient characteristics; The transient characteristics are matched with a preset non-faulty leakage current transient characteristic template; Multidimensional feature fusion analysis is performed based on the frequency components, harmonic characteristics, comparison results, intensity of transient features, duration of transient features, matching results, propagation path of zero-sequence current information in different loops, and attenuation of zero-sequence current information in different loops. Based on the results of the multidimensional feature fusion analysis, the source of the zero-sequence current is distinguished.

4. The method for hierarchical configuration of leakage protection for power operations based on risk assessment as described in claim 1, characterized in that, The step of distinguishing currents from different sources based on the characteristics of the zero-sequence current information to avoid protection actions caused by the superposition of non-fault currents includes: For each power circuit, a weak test current with a specific frequency and waveform characteristics is periodically injected; Simultaneously monitor the zero-sequence current response after the test current is injected into all power circuits; By comparing the characteristics of the monitored zero-sequence current response with those of the injected test current, the response component caused by the test current is separated. The separated response component is removed from the total zero-sequence current information to obtain zero-sequence current information that does not contain the test current response; Residual signal analysis is performed on the zero-sequence current information after removing the test current response to identify whether there are weak, intermittent signals that match the preset fault characteristics. When a weak, intermittent signal matching the preset fault characteristics is identified in the residual signal, it is determined that a real leakage fault exists.

5. A method for hierarchical configuration of leakage current protection for power operations based on risk assessment, as described in claim 1, is characterized in that... The step of distinguishing currents from different sources based on the characteristics of the zero-sequence current information to avoid protection actions caused by the superposition of non-fault currents includes: Perform time-frequency joint analysis on the zero-sequence current information to generate the time-frequency distribution of the zero-sequence current information; The energy distribution, instantaneous frequency change rate of the zero-sequence current information, and the time-varying trajectory of the harmonic components of the zero-sequence current information are extracted from the time-frequency distribution. The energy distribution, the instantaneous frequency change rate, and the trajectory of the harmonic components over time are compared with a preset electromagnetic interference characteristic time-frequency template to adapt to the distortion and drift of the characteristics on the time axis. The comparison results are corrected by combining the load change information of the power circuit and the operating status of the equipment in the power circuit; Based on the corrected comparison results, the source of the zero-sequence current is distinguished.

6. The method for hierarchical configuration of leakage current protection for power operations based on risk assessment as described in claim 1, characterized in that, The step of distinguishing currents from different sources based on the characteristics of the zero-sequence current information to avoid protection actions caused by the superposition of non-fault currents includes: Spatial correlation analysis is performed on zero-sequence current information obtained from power circuits in different work areas to identify the propagation characteristics and mutual influence relationships of the zero-sequence current information between different work areas; Based on the spatial correlation analysis results, time-series analysis is performed on the zero-sequence current information to extract the dynamic characteristics of the zero-sequence current information changing over time and the correlation at different time points. By combining the spatial correlation analysis results and the temporal analysis results, a spatiotemporal correlation diagram of zero-sequence current information is constructed; Extract the spatiotemporal difference characteristics and spatiotemporal coupling characteristics of the zero-sequence current information from the spatiotemporal correlation diagram; The spatiotemporal difference characteristics and spatiotemporal coupling characteristics are compared with a preset spatiotemporal characteristic template of non-faulty current to identify non-faulty currents with spatial differences and dynamic temporal coupling characteristics. Based on the comparison results, the source of the zero-sequence current is distinguished.

7. A method for hierarchical configuration of leakage protection for electrical work based on risk assessment, as described in claim 6, is characterized in that, The steps of extracting the spatiotemporal difference features of the zero-sequence current information and the spatiotemporal coupling characteristics of the zero-sequence current information from the spatiotemporal correlation graph include: The spatiotemporal correlation graph is decomposed at different spatiotemporal scales. To obtain the energy distribution, frequency components, and propagation path characteristics of zero-sequence current at different scales; The energy distribution, frequency components, and propagation path features are aggregated using spatiotemporal convolution operations; Extract the spatiotemporal difference characteristics and spatiotemporal coupling characteristics of the zero-sequence current information.

8. A method for hierarchical configuration of leakage current protection for power operations based on risk assessment, as described in claim 7, is characterized in that... The step of aggregating the energy distribution, the frequency components, and the propagation path features using spatiotemporal convolution operations includes: For the aforementioned spatiotemporal correlation graph, multi-scale decomposition is performed to obtain the energy distribution, frequency components, and propagation path characteristics of zero-sequence current information at different scales; Based on the energy distribution, frequency components, and propagation path characteristics of the zero-sequence current information at different scales, the size and weight of the spatiotemporal convolution kernel are dynamically adjusted. When the zero-sequence current information in a specific spatiotemporal region is detected to exhibit high-intensity energy concentration or drastic frequency changes, the spatiotemporal convolution kernel size of that region is increased, and its weights are improved. When a nonlinear correlation is detected between zero-sequence current information at different scales, the weight allocation of the spatiotemporal convolution kernel is adjusted to strengthen the aggregation of nonlinear correlation features and weaken the aggregation of linear correlation features.

9. A method for hierarchical configuration of leakage protection for power operations based on risk assessment, as described in claim 8, is characterized in that... When zero-sequence current information in a specific spatiotemporal region is detected to exhibit high-intensity energy concentration or drastic frequency changes, the steps of increasing the spatiotemporal convolution kernel size and improving its weights in that region include: For each power circuit, continuously monitor the energy intensity and frequency change rate of zero-sequence current information, and record the timestamp and duration of its instantaneous changes; When zero-sequence current information is detected to exhibit high-intensity energy concentration or drastic frequency changes in a specific spatiotemporal region, the high-speed feature sampling module is activated to collect zero-sequence current information in that region at a high sampling rate and obtain fine transient waveforms. Based on the fine transient waveform, calculate its transient energy distribution, instantaneous frequency change trajectory, and waveform distortion degree; Based on the transient energy distribution, the instantaneous frequency change trajectory, and the waveform distortion degree, the dynamic size and weight coefficient of the spatiotemporal convolution kernel are calculated in real time, wherein the dynamic size and the weight coefficient are positively correlated with the complexity and duration of the transient feature; In the spatiotemporal convolution operation, the spatiotemporal convolution kernel with the dynamic size and the weight coefficients is applied to aggregate the zero-sequence current information of the specific spatiotemporal region.

10. A risk assessment-based method for hierarchical configuration of leakage current protection in power operations, characterized in that, The system includes: The zero-sequence current information acquisition module is used to acquire zero-sequence current information of multiple power circuits. An electrical connection relationship acquisition module is used to acquire the electrical connection relationship and current distribution ratio between the multiple power circuits; The leakage current total reconstruction module is used to reconstruct the actual leakage current total of the multiple power circuits based on the zero-sequence current information, the electrical connection relationship, and the current distribution ratio. The leakage risk assessment module is used to determine whether there is a leakage risk based on the actual total leakage amount. A current source differentiation module is used to distinguish currents from different sources based on the characteristics of the zero-sequence current information, so as to avoid protection actions caused by the superposition of non-faulty currents; and The protection action triggering module is used to trigger corresponding protection actions when a leakage risk is detected.

Images