Leakage protection method and device, storage medium and leakage protection system

Abstract

The invention relates to an electric leakage protection method and device, a storage medium and an electric leakage protection system. The electric leakage protection method belongs to the technical field of electric leakage detection, and comprises the following steps: collecting electric leakage current data of a building low-voltage power distribution system through a zero-sequence current transformer and a residual current sensor; the collected data are analyzed based on an electric leakage judgment algorithm of a machine learning model, and normal leakage current and fault leakage current are distinguished; and when the fault leakage current is detected, the leakage protection action is triggered, and the building low-voltage power distribution system is linked to carry out emergency response. According to the invention, the collected data is analyzed through the electric leakage judgment algorithm based on the machine learning model, the normal leakage current and the fault leakage current are distinguished, the detection efficiency of electric leakage detection is improved, a slight electric leakage phenomenon is found, the building low-voltage power distribution system is linked to carry out emergency response, and the system safety is improved.

Description

Technical Field

[0001] This disclosure relates to the field of leakage current detection technology, and in particular to a leakage current protection method, device, storage medium and leakage current protection system. Background Technology

[0002] In existing low-voltage power distribution systems in buildings, residual current devices (RCDs) have some shortcomings. Traditional RCDs often employ relatively simple detection principles and mechanical structures, resulting in limited sensitivity. They may fail to detect minute leakage currents accurately and promptly, leading to potential safety hazards. For example, in complex electrical environments, such as those with numerous RCD systems or aging wiring, minor leakage may gradually develop into serious safety problems, which traditional RCDs fail to effectively warn of.

[0003] Furthermore, existing leakage current protection devices are not accurate enough in distinguishing between normal leakage current and fault leakage current, making them prone to malfunction. For example, during the normal operation of some electrical equipment, there will be a certain amount of normal leakage current due to the characteristics of the equipment itself, but existing protection devices may not be able to accurately identify it, thus mistakenly cutting off the circuit, affecting the normal operation of the equipment, and causing inconvenience and economic losses to users.

[0004] Moreover, most current leakage protection devices operate independently and lack effective linkage with the building's security system. When a leakage accident occurs, it is impossible to notify relevant personnel in a timely manner and take further emergency measures, which greatly reduces the timeliness and effectiveness of accident handling and may expand the scope and impact of the accident. Summary of the Invention

[0005] In view of the above, the present disclosure aims to provide a leakage current protection method, apparatus, storage medium, and leakage current protection system.

[0006] The technical solution disclosed herein is implemented as follows:

[0007] Firstly, this disclosure provides a leakage current protection method.

[0008] The leakage current protection method provided in this disclosure includes:

[0009] Leakage current data of building low-voltage power distribution systems are collected using zero-sequence current transformers and residual current sensors.

[0010] The leakage current detection algorithm based on machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current.

[0011] When a fault leakage current is detected, the leakage protection is triggered, and the building's low-voltage power distribution system is activated for emergency response.

[0012] In some embodiments, the zero-sequence current transformer employs a permalloy core and an optimized winding structure to improve the detection accuracy of minute leakage currents.

[0013] In some embodiments, the leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including:

[0014] Extract the time-domain, frequency-domain, and time-series characteristics of the leakage current;

[0015] The leakage current is classified by using support vector machines or neural networks to identify the time-domain, frequency-domain, and time-series characteristics of the leakage current, thereby identifying the fault leakage current.

[0016] In some embodiments, the neural network is a long short-term memory network used to process time-series data of leakage current and capture abnormal fluctuations.

[0017] In some embodiments, the leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including:

[0018] Monitor the temperature of electrical equipment using temperature sensors;

[0019] Based on the temperature of the electrical equipment and the data analysis results of the collected data combined with the leakage current detection algorithm, it is determined whether there is a fault leakage current in the building's low-voltage power distribution system.

[0020] In some embodiments, the linked building low-voltage power distribution system performs an emergency response, including:

[0021] Send leakage alarm information to the building's low-voltage power distribution system via serial or Ethernet interface;

[0022] The building's low-voltage power distribution system will activate audible and visual alarms, cut off the power supply, or notify management personnel based on the alarm information.

[0023] In some embodiments, before the leakage current detection algorithm based on a machine learning model analyzes the collected data, the method includes:

[0024] The leakage current data is filtered and normalized in real time to obtain processed leakage current data, which is then used for data analysis by the leakage current detection algorithm; and

[0025] The machine learning model is trained using historical data to optimize the algorithm parameters for leakage current detection.

[0026] Secondly, this disclosure provides a leakage current protection device, comprising:

[0027] The data acquisition unit is used to acquire leakage current data of the building's low-voltage power distribution system through a zero-sequence current transformer and a residual current sensor.

[0028] The data analysis unit is used to analyze the collected data based on the leakage current detection algorithm of the machine learning model, and to distinguish between normal leakage current and fault leakage current.

[0029] The leakage current protection unit is used to trigger the leakage current protection action when a fault leakage current is detected, and to link the building's low-voltage power distribution system for emergency response.

[0030] Thirdly, this disclosure provides a computer-readable storage medium having a leakage current protection program stored thereon, which, when executed by a processor, implements the leakage current protection method described in the first aspect above.

[0031] Fourthly, this disclosure provides a leakage current protection system, including a memory, a processor, and a leakage current protection program stored in the memory and executable on the processor. When the processor executes the leakage current protection program, it implements the leakage current protection method described in the first aspect above.

[0032] The leakage current protection method according to embodiments of this disclosure includes: collecting leakage current data of a building's low-voltage power distribution system using a zero-sequence current transformer and a residual current sensor; analyzing the collected data using a leakage current judgment algorithm based on a machine learning model to distinguish between normal leakage current and fault leakage current; and triggering leakage current protection action when a fault leakage current is detected, and coordinating with the building's low-voltage power distribution system for emergency response. In this application, the use of a leakage current judgment algorithm based on a machine learning model to analyze the collected data and distinguish between normal leakage current and fault leakage current improves the detection efficiency of leakage current detection, enabling the detection of minor leakage phenomena and coordinating with the building's low-voltage power distribution system for emergency response, thereby enhancing system safety.

[0033] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a leakage current protection method according to an exemplary embodiment;

[0035] Figure 2 This is a schematic diagram of a leakage current protection device structure according to an exemplary embodiment. Detailed Implementation

[0036] Embodiments of this disclosure are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0037] In existing low-voltage power distribution systems in buildings, residual current devices (RCDs) have some shortcomings. Traditional RCDs often employ relatively simple detection principles and mechanical structures, resulting in limited sensitivity. They may fail to detect minute leakage currents accurately and promptly, leading to potential safety hazards. For example, in complex electrical environments, such as those with numerous RCD systems or aging wiring, minor leakage may gradually develop into serious safety problems, which traditional RCDs fail to effectively warn of.

[0038] Furthermore, existing leakage current protection devices are not accurate enough in distinguishing between normal leakage current and fault leakage current, making them prone to malfunction. For example, during the normal operation of some electrical equipment, there will be a certain amount of normal leakage current due to the characteristics of the equipment itself, but existing protection devices may not be able to accurately identify it, thus mistakenly cutting off the circuit, affecting the normal operation of the equipment, and causing inconvenience and economic losses to users.

[0039] Moreover, most current leakage protection devices operate independently and lack effective linkage with the building's security system. When a leakage accident occurs, it is impossible to notify relevant personnel in a timely manner and take further emergency measures, which greatly reduces the timeliness and effectiveness of accident handling and may expand the scope and impact of the accident.

[0040] In view of the above situation, this disclosure provides a leakage current protection method. Figure 1 This is a flowchart illustrating a leakage current protection method according to an exemplary embodiment. Figure 1 As shown, the leakage current protection method includes:

[0041] Step 10: Collect leakage current data of the building's low-voltage power distribution system using a zero-sequence current transformer and a residual current sensor;

[0042] Step 11: The leakage current detection algorithm based on the machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current.

[0043] Step 12: When a fault leakage current is detected, the leakage protection is triggered, and the building's low-voltage power distribution system is activated for emergency response.

[0044] In this exemplary embodiment, a zero-sequence current transformer can detect zero-sequence current, and a residual current sensor can detect residual current. The machine learning model can be obtained through pre-training. The machine learning model includes a leakage current detection algorithm. The leakage current detection algorithm of the machine learning model can analyze the collected data to distinguish between normal leakage current and fault leakage current. A threshold can be set for judgment. Since the fault leakage current is greater than the normal leakage current and has different waveform characteristics, the leakage current can be determined as either normal or fault leakage current by judging its magnitude and waveform characteristics.

[0045] Meanwhile, in this application, when a fault leakage current is detected, in order to protect the circuit, the leakage protection action can be triggered, and the building's low-voltage power distribution system can be linked to carry out an emergency response, so as to improve the timeliness and effectiveness of accident handling.

[0046] In some embodiments, the zero-sequence current transformer employs a permalloy core and an optimized winding structure to improve the detection accuracy of minute leakage currents.

[0047] In this exemplary embodiment, high-quality core and winding materials can be selected, such as permalloy with high magnetic permeability, to enhance the transformer's ability to sense zero-sequence current. Simultaneously, the number of turns and wire diameter of the transformer's windings are optimized to improve its measurement accuracy and sensitivity. Its working principle is based on the law of electromagnetic induction. When a zero-sequence current appears in the three-phase circuit, i.e., when the three-phase current is unbalanced or leakage occurs, the zero-sequence current transformer will induce a corresponding potential signal, which accurately reflects the magnitude of the leakage current.

[0048] In addition to zero-sequence current transformers, other types of sensors are also used, such as residual current sensors and temperature sensors. Residual current sensors can directly measure the residual current in the circuit, complementing and verifying the measurement results of the zero-sequence current transformer. Temperature sensors are used to monitor temperature changes in electrical equipment and circuits, because leakage current is often accompanied by local temperature rise. Through data fusion analysis of multiple sensors, the leakage current situation can be judged more comprehensively and accurately.

[0049] In some embodiments, the leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including:

[0050] Extract the time-domain, frequency-domain, and time-series characteristics of the leakage current;

[0051] The leakage current is classified by using support vector machines or neural networks to identify the time-domain, frequency-domain, and time-series characteristics of the leakage current, thereby identifying the fault leakage current.

[0052] Among them, normal leakage current is usually small, relatively stable, with a waveform close to sine (or in phase with the power supply voltage), a spectrum concentrated near the fundamental wave, and changes slowly over time.

[0053] Fault leakage current: may suddenly increase, waveform may be distorted (containing a large number of harmonics, pulses, intermittent), spectrum energy distribution may be wider (rich harmonics, increased high frequency components), and change drastically or be unstable over time (such as the randomness of arc faults).

[0054] The step of classifying the time-domain, frequency-domain, and time-series characteristics of the leakage current using support vector machines or neural networks to identify fault leakage current includes:

[0055] Based on at least one of the following time-domain characteristics: effective value, peak value, waveform factor, skewness, kurtosis, and zero-crossing rate; at least one of the following frequency-domain characteristics: fundamental component amplitude / phase, harmonic component amplitude / energy, total harmonic distortion rate, specific frequency band energy, and fundamental energy percentage; and at least one of the following time-series characteristics: moving average, moving standard deviation, autocorrelation function, approximate entropy, and trend characteristics, the fault leakage current is classified and identified.

[0056] I. Time-Domain Features

[0057] These characteristics can be calculated directly from the current sample value sequence.

[0058] The criteria for determining the effective value (RMS): The effective value (RMS) of the fault leakage current is usually significantly higher than the normal leakage current (especially in cases of metallic short circuits or severe insulation aging). It is one of the most fundamental criteria for determining whether the threshold has been exceeded.

[0059] Note: Some normal startup processes or momentary disturbances may also cause a temporary increase in RMS, which needs to be judged in conjunction with other characteristics.

[0060] Peak Value / Peak-to-Peak Value:

[0061] Judgment criteria: Faults (especially arc faults and pulse faults) may cause an abnormal increase in current peak value, far exceeding the expected peak value range of normal leakage current.

[0062] Form Factor: The criterion for determining RMS / |Average| is as follows: Normal leakage current is close to a sine wave, with a crest factor of approximately 1.4. Fault currents (such as arc faults) have severely distorted waveforms and may exhibit spikes, resulting in a significantly increased crest factor.

[0063] The criterion for determining skewness is to measure the asymmetry of the waveform distribution. Normal leakage current waveforms are symmetrical, with skewness close to 0. Certain faults (such as asymmetrical breakdown or half-wave rectification effect) may cause the waveform to deviate to one side, increasing the absolute value of the skewness.

[0064] Kurtosis is determined by measuring the sharpness or flatness of the waveform distribution (relative to a normal distribution). Normal leakage current has a kurtosis close to 3 (the kurtosis value of a normal distribution). Arc faults and other abnormalities can generate numerous instantaneous spikes, significantly increasing the kurtosis (>4 or higher), indicating the presence of abnormal spikes in the waveform.

[0065] The criteria for determining the number of local maximum / minimum values ​​are as follows: Normal leakage current waveforms are smooth, with fewer and more regular local extreme points. Arc fault currents are highly unstable, generating numerous high-frequency oscillations and random spikes, leading to a significant increase in the number of local extreme points.

[0066] The criteria for determining the zero-crossing rate are: a stable zero-crossing rate for normal leakage current (approximately 100 times / second for a 50Hz system). High-frequency oscillations or noise interference (common in faults) will significantly increase the zero-crossing rate.

[0067] II. Frequency Domain Features

[0068] These features are calculated by performing a Fast Fourier Transform (FFT) or Wavelet Transform on the current signal.

[0069] Fundamental component amplitude / phase:

[0070] Judgment criteria: An excessively large fundamental frequency amplitude is a direct indicator of a serious fault. Abnormal phase changes may also indicate a specific fault type (such as changes in capacitive / inductive leakage current).

[0071] Harmonic component magnitude / energy: (with particular attention to odd harmonics such as the 3rd, 5th, and 7th).

[0072] Judgment criteria: Normal leakage current has very low harmonic content. Faults (especially arc faults and semiconductor device faults) will generate abundant harmonics, and a significant increase in the amplitude or energy of specific harmonics (such as the 3rd and 5th harmonics) is a strong indication of a fault.

[0073] Total Harmonic Distortion (THD):

[0074] Judgment criterion: THD = sqrt(∑(harmonic amplitude²)) / fundamental amplitude. Normal leakage current has very low THD. Fault current (especially nonlinear faults such as arcing) will cause a significant increase in THD. This is a key characteristic distinguishing faults (especially arcing) from normal conditions.

[0075] Energy in Specific Frequency Bands:

[0076] Judgment criteria: Dividing different frequency bands (e.g., fundamental frequency band, low-frequency harmonic band, high-frequency noise band). Fault current often has a significantly higher energy proportion in the high-frequency band (e.g., >2kHz or higher) than normal current (high-frequency noise generated by the arc). The energy proportion in the low-frequency harmonic band may also be increased.

[0077] Fundamental Energy Ratio:

[0078] Judgment criterion: Fundamental frequency energy / Total energy. Normally, leakage current energy is highly concentrated in the fundamental frequency. During a fault, the energy is dispersed into harmonics and noise, causing this ratio to decrease.

[0079] III. Time-Series Features

[0080] These features focus on the characteristics of how the current signal evolves over time.

[0081] The moving average (MA) is used to smooth out short-term fluctuations and observe long-term trends. The MA of a fault current may show a continuous increase, a step-like rise, or sharp fluctuations, unlike the steady or slow changes of normal current.

[0082] The moving standard deviation (MSD) / moving variance is determined by measuring the volatility of a signal within a sliding window. Normal leakage current exhibits small and stable volatility. Faults (especially intermittent arcing and unstable faults) will cause a significant increase in volatility (MSD).

[0083] The criteria for determining the characteristics of the autocorrelation function are: calculating the correlation between the signal and its own delayed version. Normal alternating current has strong periodicity, and the autocorrelation function will peak at the fundamental period and its integer multiples of the delay. Faults (especially those that disrupt periodicity, such as random arcs) weaken this periodicity, leading to a decrease in the autocorrelation peak and faster decay.

[0084] Approximate Entropy (ApEn) / Sample Entropy (SampEn):

[0085] Judgment criteria: Measuring the complexity and irregularity of the time series. Normal leakage current is regular, highly periodic, and has a low entropy value. Fault current (especially arcing) has a high degree of randomness and complexity, and its entropy value will increase significantly.

[0086] Trend Features: (e.g., linear regression slope) Criteria for determining trend features: Calculate the linear regression slope of current values ​​within a window. A continuously rising slope may indicate a slow increase in leakage current due to gradual insulation degradation (early failure).

[0087] A single characteristic is usually insufficient for reliable determination: for example, a high RMS may be a fault, but it may also be a normal high-load start-up; a high crest factor may be an electric arc, but it may also be a transient disturbance.

[0088] The role of the model: The core value of models such as SVM or NN lies in learning the complex nonlinear combination relationships between these features, thereby finding the optimal separating boundary (SVM) or mapping function (NN) in the multidimensional feature space.

[0089] Example of a combination of characteristics of a typical failure mode:

[0090] Severe short circuit / ground fault: extremely high RMS, high peak value, huge fundamental amplitude, THD may not be high (close to pure sine distortion).

[0091] Series / parallel arc faults: RMS may be moderate or variable, extremely high crest factor, extremely high kurtosis, numerous local extrema, significantly increased THD (especially mid-to-low frequency harmonics), significantly increased high frequency band energy, significantly increased sample entropy / approximate entropy, and increased moving standard deviation.

[0092] Insulation aging leads to a slow increase in leakage current: RMS rises slowly and continuously (trend characteristic slope > 0), while other characteristics may not change much.

[0093] The decision process (within the model) calculates the aforementioned time-domain, frequency-domain, and time-series features of the input leakage current signal to form a feature vector.

[0094] Input the feature vectors into the trained SVM or NN model.

[0095] The distance from the feature vector to the optimal classification hyperplane can be calculated using SVM, or the feature vector can be mapped to a high-dimensional space using a kernel function for judgment.

[0096] The NN feature vectors are used to perform calculations and nonlinear transformations through a multi-layer network.

[0097] The model outputs a classification result (faulty / normal) and / or the probability of belonging to each category.

[0098] Data quality and sampling rate: High-frequency features (such as crest factors, local extrema, and high-frequency energy) require a sufficiently high sampling rate to be accurately captured. Anti-aliasing filtering is also important.

[0099] Feature selection: Not all features are equally effective or independent. Using correlation analysis, principal component analysis, recursive feature elimination, or feature importance-based methods (such as tree models) to select the most discriminative subset of features can improve model efficiency and generalization ability.

[0100] Feature Scaling: SVMs (especially distance-based kernels such as RBF) and NNs are often sensitive to feature scale. Standardization or normalization is a necessary preprocessing step.

[0101] Labeled data: A large number of accurately labeled "normal" and "fault" (different types) leakage current samples are needed for training and validating the model.

[0102] Diverse fault types: The model needs to be able to distinguish between different types of faults (arc, short circuit, insulation aging, etc.) and different normal operating conditions (different loads, startup transients, etc.), which requires the training data to cover these situations.

[0103] Real-time requirements: Select features and models with appropriate computational complexity based on the application scenario (e.g., circuit breaker protection requires extremely fast response, while monitoring systems can be slightly slower).

[0104] The core of identifying fault leakage current using SVM or NN lies in constructing a multi-dimensional feature space that can effectively distinguish between normal and fault modes. The key is the combined use of time-domain features (RMS, crest factor, kurtosis, local extrema), frequency-domain features (THD, high-frequency energy, specific harmonic amplitudes), and time-series features (moving standard deviation, sample entropy, autocorrelation characteristics). By learning the complex relationships between these features, the model can determine faults more robustly and accurately than a single threshold or simple rule. The final determination is the result of a comprehensive decision based on all feature values ​​calculated by the model from the input signal. Feature engineering (selection, construction, scaling) and high-quality labeled data are fundamental to the model's success.

[0105] In this exemplary embodiment, machine learning algorithms, such as Support Vector Machines (SVM) and neural networks, can be used to classify and identify the extracted feature parameters. Through system "self-learning," a large number of normal leakage current values ​​and waveforms for common electrical devices can be added to the algorithm database. The system can deeply understand the comprehensive superimposed current waveform of normal leakage current and compare it with the current waveform collected by sensors in real time. The algorithm is trained and optimized through massive amounts of experimental data and practical cases, enabling it to accurately distinguish between normal leakage current and fault leakage current. When multiple electrical devices are connected to the same circuit simultaneously, their normal leakage currents will be superimposed to form the total leakage current. According to the principle of linear superposition, the leakage currents of multiple devices can be simply added to obtain the total leakage current. The formula is as follows: Itotal = I1 + I2 + I3 + ... + In; where: Itotal is the total leakage current of all devices; I1, I2, I3, ..., In are the normal leakage currents of each device. The superposition of normal leakage currents is based on the leakage current value of the device under normal operating conditions. If a device malfunctions, its leakage current may suddenly increase, forming a fault leakage current. At this time, the total leakage current will no longer follow a simple linear superposition law.

[0106] Among them, deep learning models (such as neural networks) have powerful nonlinear modeling capabilities and can automatically extract complex features from large amounts of data, making them suitable for processing variable electrical system data.

[0107] For example, normal leakage current is generally less than or equal to 1mA, while fault leakage current is greater than or equal to 30mA; the waveform of normal leakage current is a sine wave, in sync with the power supply frequency; fault leakage current may contain high-frequency harmonics or DC components, and so on. Deep learning models (such as neural networks) can be used to analyze leakage current characteristics, thereby distinguishing between normal and fault leakage current.

[0108] Among them, current transformers, leakage current sensors and other equipment are installed at key nodes of the electrical system (such as distribution boxes, equipment inlets, etc.) to collect leakage current data in real time.

[0109] Choose an appropriate data acquisition frequency based on the application scenario. For real-time monitoring, a acquisition frequency of once per second or higher is recommended; for long-term analysis, the acquisition frequency can be appropriately reduced.

[0110] The collected data is stored in a database or cloud platform to ensure data security and accessibility.

[0111] Use filters (such as low-pass filters and band-pass filters) to remove high-frequency noise from the signal while retaining useful leakage current information.

[0112] If missing values ​​exist in the data, interpolation or mean imputation methods can be used to complete them. Normalize the leakage current data to the range [0,1] or [-1,1] to facilitate model training and convergence. Extract statistical features of the leakage current, such as mean, variance, peak value, and kurtosis, to reflect the overall characteristics of the current. Transform the leakage current from the time domain to the frequency domain using Fourier transform or wavelet transform, extract the energy distribution of different frequency components, and identify harmonic components. Extract time-series features of the leakage current, such as autocorrelation coefficient, partial autocorrelation coefficient, and moving average, to capture the dynamic trend of current changes. Combine other sensor data (such as temperature, humidity, vibration, etc.) to construct a multivariate feature vector to improve the model's prediction accuracy.

[0113] In some embodiments, the neural network is a long short-term memory network used to process time-series data of leakage current and capture abnormal fluctuations.

[0114] In this exemplary embodiment, the deep learning model chosen is a Long Short-Term Memory (LSTM) network, a special type of recurrent neural network (RNN) that excels at processing time-series data. For time-dependent data such as leakage current, LSTM can effectively capture the trend of current changes over time and identify abnormal current fluctuations. It can handle long-term dependencies and is suitable for predicting future leakage current values ​​or detecting anomalies. However, it requires a long training time and a large amount of data.

[0115] In some embodiments, the leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including:

[0116] Monitor the temperature of electrical equipment using temperature sensors;

[0117] Based on the temperature of the electrical equipment and the data analysis results of the collected data combined with the leakage current detection algorithm, it is determined whether there is a fault leakage current in the building's low-voltage power distribution system.

[0118] In this exemplary embodiment, the parameters in the algorithm are initialized according to different electrical devices and application scenarios. For example, different leakage current thresholds are set for electrical devices with different power ratings; and the alarm threshold of the temperature sensor is adjusted for different ambient temperature ranges. Furthermore, during device operation, the parameters can be dynamically adjusted according to actual conditions to adapt to constantly changing operating conditions. For example, when the temperature rises abnormally and the leakage current exceeds the threshold, a fault leakage is determined.

[0119] Integrate the residual current device (RCD) with the building's existing security system, configuring communication parameters and linkage rules. During the integration process, rigorous testing must be conducted to ensure that both systems can communicate and work collaboratively.

[0120] A comprehensive test was conducted on the entire intelligent leakage current protection system, including sensor performance testing, algorithm accuracy testing, and linkage function testing. Under different load conditions and leakage current simulations, the system's various functions were verified to ensure they met design requirements. For example, different degrees of leakage current faults were simulated to check whether the system could accurately detect, judge, and promptly activate corresponding emergency measures.

[0121] In some embodiments, the linked building low-voltage power distribution system performs an emergency response, including:

[0122] Send leakage alarm information to the building's low-voltage power distribution system via serial or Ethernet interface;

[0123] The building's low-voltage power distribution system will activate audible and visual alarms, cut off the power supply, or notify management personnel based on the alarm information.

[0124] In this exemplary embodiment, a standard communication interface, such as RS485 or Ethernet interface, is provided in the leakage current protection device to facilitate data communication with the building's security system. A common communication protocol is used to ensure the stability and compatibility of data transmission.

[0125] In this exemplary embodiment, information interaction and collaborative operation are as follows: When the leakage current protection device detects a leakage accident, it immediately sends alarm information to the security system via the communication interface. The alarm information includes detailed data such as the location of the leakage, the magnitude of the leakage current, and the time of occurrence. After receiving the alarm information, the security system activates corresponding emergency measures according to preset rules and strategies, such as issuing audible and visual alarm signals to notify relevant personnel to rush to the scene; automatically cutting off the power supply to the relevant area to prevent the accident from escalating; and uploading the alarm information to the property management center or remote monitoring platform so that managers can promptly grasp the accident situation and make further decisions.

[0126] System Integration and Testing: A comprehensive system integration and testing process is conducted between the leakage current protection device and the security system to ensure accurate information exchange and stable, reliable collaborative operation. Experiments are performed under different simulated leakage current scenarios to verify the effectiveness and timeliness of the linkage function.

[0127] In some embodiments, before the leakage current detection algorithm based on a machine learning model analyzes the collected data, the method includes:

[0128] The leakage current data is filtered and normalized in real time to obtain processed leakage current data, which is then used for data analysis by the leakage current detection algorithm; and

[0129] The machine learning model is trained using historical data to optimize the algorithm parameters for leakage current detection.

[0130] In this exemplary embodiment, the collected data is divided into a training set, a validation set, and a test set. Typically, the training set accounts for 70%, the validation set accounts for 15%, and the test set accounts for 15%.

[0131] Choose an appropriate loss function based on the task type. For regression tasks (such as predicting leakage current), commonly used loss functions include mean squared error (MSE) and mean squared logarithmic error (MSLE); for classification tasks (such as normal / abnormal classification), commonly used loss functions include cross-entropy loss.

[0132] Use gradient descent methods (such as Adam and RMSprop) to optimize model parameters, ensuring that the model can converge quickly and achieve good performance.

[0133] By using methods such as grid search, random search, or Bayesian optimization, the hyperparameters of the model (such as learning rate, batch size, number of hidden layer nodes, etc.) can be adjusted to find the optimal model configuration.

[0134] Use k-fold cross-validation (such as 5-fold or 10-fold) to evaluate the model's generalization ability and ensure that the model does not overfit.

[0135] Choose the appropriate performance metric based on the task type. For regression tasks, commonly used metrics include mean squared error (MSE) and mean absolute error (MAE); for classification tasks, commonly used metrics include accuracy, recall, and F1 score.

[0136] The performance of the model can be intuitively evaluated by plotting charts such as confusion matrix, ROC curve, and PR curve.

[0137] The trained model can be deployed to edge computing devices or cloud servers to achieve real-time monitoring and early warning. When an abnormal leakage current is detected, the system can automatically issue an alarm to notify relevant personnel for inspection and handling.

[0138] By applying the model to historical data, we can analyze the long-term operating status of equipment, identify potential problems and trends, and provide a basis for equipment maintenance.

[0139] By combining Internet of Things (IoT) technology with automated control systems, intelligent equipment maintenance can be achieved. For example, when a model predicts that a piece of equipment is about to fail, the system can automatically schedule maintenance tasks to avoid equipment downtime.

[0140] This disclosure provides a leakage current protection device. Figure 2 This is a schematic diagram of a leakage current protection device structure according to an exemplary embodiment. Figure 2 As shown, the leakage current protection device includes:

[0141] Data acquisition unit 20 is used to acquire leakage current data of building low-voltage power distribution system through zero-sequence current transformer and residual current sensor;

[0142] Data analysis unit 21 is used to analyze the collected data by a leakage current judgment algorithm based on a machine learning model, and to distinguish between normal leakage current and fault leakage current.

[0143] The leakage current protection unit 22 is used to trigger the leakage current protection action when a fault leakage current is detected, and to link the building's low-voltage power distribution system for emergency response.

[0144] In this exemplary embodiment, a zero-sequence current transformer can detect zero-sequence current, and a residual current sensor can detect residual current. The machine learning model can be obtained through pre-training. The machine learning model includes a leakage current detection algorithm. The leakage current detection algorithm of the machine learning model can analyze the collected data to distinguish between normal leakage current and fault leakage current. A threshold can be set for judgment. Since the fault leakage current is greater than the normal leakage current and has different waveform characteristics, the leakage current can be determined as either normal or fault leakage current by judging its magnitude and waveform characteristics.

[0145] Meanwhile, in this application, when a fault leakage current is detected, in order to protect the circuit, the leakage protection action can be triggered, and the building's low-voltage power distribution system can be linked to carry out an emergency response, so as to improve the timeliness and effectiveness of accident handling.

[0146] In some embodiments, the data analysis unit 21 is used for

[0147] Extract the time-domain, frequency-domain, and time-series characteristics of the leakage current;

[0148] The leakage current is classified by using support vector machines or neural networks to identify the time-domain, frequency-domain, and time-series characteristics of the leakage current, thereby identifying the fault leakage current.

[0149] In this exemplary embodiment, machine learning algorithms, such as Support Vector Machines (SVM) and neural networks, can be used to classify and identify the extracted feature parameters. Through system "self-learning," a large number of normal leakage current values ​​and waveforms for common electrical devices can be added to the algorithm database. The system can deeply understand the comprehensive superimposed current waveform of normal leakage current and compare it with the current waveform collected by sensors in real time. The algorithm is trained and optimized through massive amounts of experimental data and practical cases, enabling it to accurately distinguish between normal leakage current and fault leakage current. When multiple electrical devices are connected to the same circuit simultaneously, their normal leakage currents will be superimposed to form the total leakage current. According to the principle of linear superposition, the leakage currents of multiple devices can be simply added to obtain the total leakage current. The formula is as follows: Itotal = I1 + I2 + I3 + ... + In; where: Itotal is the total leakage current of all devices; I1, I2, I3, ..., In are the normal leakage currents of each device. The superposition of normal leakage currents is based on the leakage current value of the device under normal operating conditions. If a device malfunctions, its leakage current may suddenly increase, forming a fault leakage current. At this time, the total leakage current will no longer follow a simple linear superposition law.

[0150] Among them, deep learning models (such as neural networks) have powerful nonlinear modeling capabilities and can automatically extract complex features from large amounts of data, making them suitable for processing variable electrical system data.

[0151] Among them, current transformers, leakage current sensors and other equipment are installed at key nodes of the electrical system (such as distribution boxes, equipment inlets, etc.) to collect leakage current data in real time.

[0152] Choose an appropriate data acquisition frequency based on the application scenario. For real-time monitoring, a acquisition frequency of once per second or higher is recommended; for long-term analysis, the acquisition frequency can be appropriately reduced.

[0153] The collected data is stored in a database or cloud platform to ensure data security and accessibility.

[0154] Use filters (such as low-pass filters and band-pass filters) to remove high-frequency noise from the signal while retaining useful leakage current information.

[0155] If missing values ​​exist in the data, interpolation or mean imputation methods can be used to complete them. Normalize the leakage current data to the range [0,1] or [-1,1] to facilitate model training and convergence. Extract statistical features of the leakage current, such as mean, variance, peak value, and kurtosis, to reflect the overall characteristics of the current. Transform the leakage current from the time domain to the frequency domain using Fourier transform or wavelet transform, extract the energy distribution of different frequency components, and identify harmonic components. Extract time-series features of the leakage current, such as autocorrelation coefficient, partial autocorrelation coefficient, and moving average, to capture the dynamic trend of current changes. Combine other sensor data (such as temperature, humidity, vibration, etc.) to construct a multivariate feature vector to improve the model's prediction accuracy.

[0156] In some embodiments, the neural network is a long short-term memory network used to process time-series data of leakage current and capture abnormal fluctuations.

[0157] In this exemplary embodiment, the deep learning model chosen is a Long Short-Term Memory (LSTM) network, a special type of recurrent neural network (RNN) that excels at processing time-series data. For time-dependent data such as leakage current, LSTM can effectively capture the trend of current changes over time and identify abnormal current fluctuations. It can handle long-term dependencies and is suitable for predicting future leakage current values ​​or detecting anomalies. However, it requires a long training time and a large amount of data.

[0158] In some embodiments, the data analysis unit 21 is used for

[0159] Monitor the temperature of electrical equipment using temperature sensors;

[0160] Based on the temperature of the electrical equipment and the data analysis results of the collected data combined with the leakage current detection algorithm, it is determined whether there is a fault leakage current in the building's low-voltage power distribution system.

[0161] In this exemplary embodiment, the parameters in the algorithm are initialized according to different electrical devices and application scenarios. For example, different leakage current thresholds are set for electrical devices with different power ratings; and the alarm threshold of the temperature sensor is adjusted for different ambient temperature ranges. Furthermore, during device operation, the parameters can be dynamically adjusted according to actual conditions to adapt to constantly changing operating conditions.

[0162] Integrate the residual current device (RCD) with the building's existing security system, configuring communication parameters and linkage rules. During the integration process, rigorous testing must be conducted to ensure that both systems can communicate and work collaboratively.

[0163] A comprehensive test was conducted on the entire intelligent leakage current protection system, including sensor performance testing, algorithm accuracy testing, and linkage function testing. Under different load conditions and leakage current simulations, the system's various functions were verified to ensure they met design requirements. For example, different degrees of leakage current faults were simulated to check whether the system could accurately detect, judge, and promptly activate corresponding emergency measures.

[0164] In some embodiments, the leakage protection unit 22 is used for

[0165] Send leakage alarm information to the building's low-voltage power distribution system via serial or Ethernet interface;

[0166] The building's low-voltage power distribution system will activate audible and visual alarms, cut off the power supply, or notify management personnel based on the alarm information.

[0167] In this exemplary embodiment, a standard communication interface, such as RS485 or Ethernet interface, is provided in the leakage current protection device to facilitate data communication with the building's security system. A common communication protocol is used to ensure the stability and compatibility of data transmission.

[0168] In this exemplary embodiment, information interaction and collaborative operation are as follows: When the leakage current protection device detects a leakage accident, it immediately sends alarm information to the security system via the communication interface. The alarm information includes detailed data such as the location of the leakage, the magnitude of the leakage current, and the time of occurrence. After receiving the alarm information, the security system activates corresponding emergency measures according to preset rules and strategies, such as issuing audible and visual alarm signals to notify relevant personnel to rush to the scene; automatically cutting off the power supply to the relevant area to prevent the accident from escalating; and uploading the alarm information to the property management center or remote monitoring platform so that managers can promptly grasp the accident situation and make further decisions.

[0169] System Integration and Testing: A comprehensive system integration and testing process is conducted between the leakage current protection device and the security system to ensure accurate information exchange and stable, reliable collaborative operation. Experiments are performed under different simulated leakage current scenarios to verify the effectiveness and timeliness of the linkage function.

[0170] This disclosure provides a computer-readable storage medium storing a leakage current protection program thereon, which, when executed by a processor, implements the leakage current protection method described in the first aspect above.

[0171] This disclosure provides a leakage current protection system, including a memory, a processor, and a leakage current protection program stored in the memory and executable on the processor. When the processor executes the leakage current protection program, it implements the leakage current protection method described in the first aspect above.

[0172] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0173] It should be understood that various parts of this disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0174] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0175] In the description of this disclosure, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this disclosure and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this disclosure.

[0176] Furthermore, the terms "first," "second," etc., used in the embodiments of this disclosure are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this disclosure can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this disclosure, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly specified in the embodiments.

[0177] In this disclosure, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing," etc., appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure based on the specific implementation.

[0178] In this disclosure, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0179] Although embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A leakage current protection method, characterized in that, include: Leakage current data of building low-voltage power distribution systems are collected using zero-sequence current transformers and residual current sensors. The leakage current detection algorithm based on machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current. When a fault leakage current is detected, the leakage protection is triggered, and the building's low-voltage power distribution system is activated for emergency response.

2. The method according to claim 1, characterized in that, The zero-sequence current transformer uses a permalloy core and an optimized winding structure to improve the detection accuracy of minute leakage currents.

3. The method according to claim 2, characterized in that, The leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including: Extract the time-domain, frequency-domain, and time-series characteristics of the leakage current; The leakage current is classified by using support vector machines or neural networks to identify the time-domain, frequency-domain, and time-series characteristics of the leakage current, thereby identifying the fault leakage current.

4. The method according to claim 3, characterized in that, The neural network is a long short-term memory network, used to process time-series data of leakage current and capture abnormal fluctuations.

5. The method according to claim 1, characterized in that, The leakage current detection algorithm based on a machine learning model analyzes the collected data to distinguish between normal leakage current and fault leakage current, including: Monitor the temperature of electrical equipment using temperature sensors; Based on the temperature of the electrical equipment and the data analysis results of the collected data combined with the leakage current detection algorithm, it is determined whether there is a fault leakage current in the building's low-voltage power distribution system.

6. The method according to claim 1, characterized in that, The linked building low-voltage power distribution system shall conduct an emergency response, including: Send leakage alarm information to the building's low-voltage power distribution system via serial or Ethernet interface; The building's low-voltage power distribution system will activate audible and visual alarms, cut off the power supply, or notify management personnel based on the alarm information.

7. The method according to claim 1, characterized in that, Before analyzing the collected data, the leakage current detection algorithm based on a machine learning model includes the following methods: The leakage current data is filtered and normalized in real time to obtain processed leakage current data, which is then used for data analysis by the leakage current detection algorithm; and The machine learning model is trained using historical data to optimize the algorithm parameters for leakage current detection.

8. A leakage current protection device, characterized in that, include: The data acquisition unit is used to acquire leakage current data of the building's low-voltage power distribution system through a zero-sequence current transformer and a residual current sensor. The data analysis unit is used to analyze the collected data based on the leakage current detection algorithm of the machine learning model, and to distinguish between normal leakage current and fault leakage current. The leakage current protection unit is used to trigger the leakage current protection action when a fault leakage current is detected, and to link the building's low-voltage power distribution system for emergency response.

9. A computer-readable storage medium, characterized in that, It stores a leakage current protection program, which, when executed by the processor, implements the leakage current protection method according to any one of claims 1-7.

10. A leakage current protection system, characterized in that, The device includes a memory, a processor, and a leakage current protection program stored in the memory and executable on the processor. When the processor executes the leakage current protection program, it implements the leakage current protection method according to any one of claims 1-7.

Images