Electric arc detection method, device and equipment of energy storage system and medium

Abstract

The invention discloses an arc detection method, device and equipment of an energy storage system and a medium, and the method comprises the steps: obtaining the current data of the energy storage system, and the current data comprises DC data and AC data; working condition characteristics are determined according to the direct current data, alternating current frequency domain characteristics and alternating current time domain characteristics are determined according to the alternating current data, and the working condition characteristics are used for indicating the current working condition of the energy storage system; and inputting the working condition characteristics, the alternating current frequency domain characteristics and the alternating current time domain characteristics into an arc detection model to obtain an arc detection result. The accuracy of arc detection can be improved.

Description

Technical Field

[0001] This application relates to the field of motor control technology, specifically to an arc detection method, device, equipment, and medium for an energy storage system. Background Technology

[0002] With the rapid development of high-capacity, long-duration energy storage technologies, the DC-side voltage and current levels of energy storage systems have significantly increased, along with the number of individual battery cells and electrical connection points. This change has significantly increased the probability of electric arcs (i.e., arcing) occurring, and also increased the difficulty of their detection. An electric arc refers to the discharge phenomenon formed when current breaks down air through an unexpected path (such as air or an insulating medium), creating a plasma channel with high temperature and intense light.

[0003] Arc detection in energy storage systems faces complex and variable operating conditions. The DC current varies widely; for example, the DC arc current in photovoltaic energy storage systems typically does not exceed 30A, while in pure energy storage systems it can exceed 200A. This increase in DC current not only enhances the arc's ignition capability but also significantly amplifies the frequency domain noise of the current signal. Furthermore, energy storage systems operate under various conditions, such as charging / discharging and grid-connected / off-grid operation. The current waveform characteristics differ significantly under these different conditions, leading to inconsistent arc signal performance and further increasing the difficulty of arc identification. Summary of the Invention

[0004] This application provides an arc detection method, apparatus, device, and medium for an energy storage system to improve the accuracy of arc detection.

[0005] In a first aspect, this application provides an arc detection method for an energy storage system, the method comprising:

[0006] Acquire current data from the energy storage system, including both DC and AC data;

[0007] Operating condition characteristics are determined based on DC data, and AC frequency domain characteristics and AC time domain characteristics are determined based on AC data. The operating condition characteristics are used to indicate the current operating condition of the energy storage system.

[0008] The operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics are input into the arc detection model to obtain the arc detection results.

[0009] In some embodiments of this application, the arc detection model includes a first branch, a second branch, and a third branch;

[0010] The operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics are input into the arc detection model to obtain the arc detection results, including:

[0011] The first feature is obtained by processing the operating condition characteristics and AC time domain characteristics through the first branch;

[0012] The second feature is obtained by processing the AC frequency domain features through the second branch;

[0013] The third feature is obtained by processing the time-domain features of the AC signal through the third branch;

[0014] The arc detection result is determined based on the first feature, the second feature, and the third feature.

[0015] In some embodiments of this application, the operating condition characteristics include DC signal characteristics and charge / discharge state. Determining the operating condition characteristics based on DC data includes:

[0016] The DC data is divided into multiple segments to obtain the characteristics of the DC signal; and the charging and discharging state is determined based on the direction indicated by the DC data.

[0017] The first branch processes the operating condition characteristics and AC time-domain characteristics to obtain the first feature, including:

[0018] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, and AC time-domain characteristics through the first branch.

[0019] In some embodiments of this application, the method further includes:

[0020] Get and log off-network status;

[0021] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, and AC time-domain characteristics through the first branch, including:

[0022] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, grid connection and disconnection states, and AC time domain characteristics through the first branch.

[0023] In some embodiments of this application, the arc detection result includes a first score value and a second score value. The first score value is used to indicate the confidence level that an arc has occurred in the energy storage system, and the second score value is used to indicate the confidence level that no arc has occurred in the energy storage system.

[0024] After obtaining the arc detection results, the method also includes:

[0025] Based on the mapping relationship between the grid connection and disconnection status and the score increment, obtain the first score increment or the second score increment corresponding to the grid connection and disconnection status;

[0026] The first score value is adjusted based on the first score increment; or the second score value is adjusted based on the second score increment.

[0027] The first score is compared with the adjusted second score, or the adjusted first score is compared with the second score, and the final arc detection result is determined based on the comparison result.

[0028] In some embodiments of this application, the AC data includes an AC time-domain signal, and determining the AC frequency-domain characteristics and AC time-domain characteristics based on the AC data includes:

[0029] The number of sampling points with current values ​​greater than the current threshold and the number of sampling points with current values ​​lower than the current threshold in the AC time-domain signal are counted respectively to obtain the first time-domain statistical feature and the second time-domain statistical feature. The AC time-domain feature includes the AC time-domain signal, the first time-domain statistical feature and the second time-domain statistical feature.

[0030] The first branch processes the operating condition characteristics and AC time-domain characteristics to obtain the first feature, including:

[0031] The first feature is obtained by processing the operating condition characteristics, the first time-domain statistical characteristics, and the second time-domain statistical characteristics through the first branch;

[0032] The third branch processes the time-domain features of the AC signal to obtain the third feature, which includes:

[0033] The third feature is obtained by processing the AC time-domain signal through the third branch.

[0034] In some embodiments of this application, both the second branch and the third branch include multiple convolutional blocks;

[0035] The second branch processes the AC frequency domain features to obtain the second feature, including:

[0036] The AC frequency domain features are convolved through multiple convolutional blocks in the second branch to obtain the second feature;

[0037] The third branch processes the time-domain features of the AC signal to obtain the third feature, which includes:

[0038] The third feature is obtained by convolving the temporal features of the communication through multiple convolutional blocks in the third branch.

[0039] In some embodiments of this application, the arc detection model is obtained in the following manner:

[0040] Acquire current sample data and corresponding arc label information;

[0041] The pre-built arc detection model is trained based on current sample data and arc label information to obtain an initial arc detection model.

[0042] The initial arc detection model is fine-tuned using both on-grid and off-grid data to obtain the final arc detection model.

[0043] Secondly, this application also provides an arc detection device for an energy storage system, the device comprising:

[0044] The data acquisition module is used to acquire the current data of the energy storage system, which includes both DC and AC data.

[0045] The feature determination module is used to determine the operating condition characteristics based on DC data and the AC frequency domain characteristics and AC time domain characteristics based on AC data. The operating condition characteristics are used to indicate the current operating condition of the energy storage system.

[0046] The detection module is used to input operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics into the arc detection model to obtain arc detection results.

[0047] Thirdly, this application also provides an electronic device, including a memory and a processor; the memory stores a computer program, and the processor runs the computer program in the memory to perform the operations in the arc detection method of the energy storage system provided in the first aspect.

[0048] Fourthly, this application also provides a storage medium storing a plurality of instructions adapted for loading by a processor to execute the steps in the arc detection method of the energy storage system provided in the first aspect. For example, the storage medium is a computer-readable storage medium.

[0049] Through one or more embodiments of the above embodiments in this application, at least the following technical effects can be achieved:

[0050] The arc detection method, apparatus, equipment, and medium for energy storage systems provided in this application include an arc detection method that acquires current data from the DC and AC sides of the energy storage system, and then extracts operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics from these data. Operating condition characteristics reflect whether the energy storage system is currently charging, discharging, or unloaded. Current fluctuation patterns differ significantly under different operating conditions, and the arc detection model can dynamically adjust feature thresholds accordingly to reduce misjudgments caused by changes in operating conditions and improve the model's adaptability and robustness. Frequency domain characteristics reveal phenomena such as changes in harmonic content and increased total harmonic distortion caused by arc occurrence, exhibiting good global identification capabilities. Time domain characteristics capture transient pulse behaviors during arc discharge, such as current envelope fluctuations, pulse width changes, and zero-crossing offsets, thus effectively supplementing the frequency domain characteristics. The combination of operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics enhances the arc detection model's ability to identify different types of arcs and improves detection accuracy. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic flowchart of an arc detection method for an energy storage system provided in an embodiment of this application;

[0053] Figure 2 This is a schematic diagram of the structure of the arc detection model provided in the embodiments of this application;

[0054] Figure 3 This is a schematic diagram of the arc detection model training process provided in the embodiments of this application;

[0055] Figure 4 This is a schematic diagram of the structure of the arc detection device for the energy storage system provided in the embodiments of this application;

[0056] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0059] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0060] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not preclude applicability to or configuration to devices performing additional tasks or steps. Furthermore, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more conditions or values ​​may in practice be based on additional conditions or values ​​beyond those conditions.

[0061] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0062] Arcing energy in energy storage systems is high, and arc extinguishing time is short. According to relevant regulations, the DC arc extinguishing time of photovoltaic systems must not exceed 2.5 seconds. Within the same duration, the arcing energy of energy storage systems is approximately 20 times that of photovoltaic systems. Such high energy accumulation in a short period of time can easily induce thermal runaway. Therefore, high-speed and accurate arc detection and rapid arc extinguishing response are essential to effectively ensure the safe and reliable operation of the system.

[0063] To improve the accuracy of arc detection, the following describes the arc detection method, apparatus, equipment, and medium for energy storage systems provided in the embodiments of this application, with reference to the accompanying drawings.

[0064] like Figure 1 As shown in the figure, this application provides an arc detection method for an energy storage system, which includes the following steps:

[0065] S101, acquires current data of the energy storage system, including DC data and AC data.

[0066] Among them, DC data can be directly obtained from the output end of the energy storage system, reflecting the changes in current amplitude and polarity during the charging and discharging process; AC data comes from the inverter output side of the energy storage system, including multiphase AC current waveforms.

[0067] S102 determines the operating condition characteristics based on DC data and the AC frequency domain characteristics and AC time domain characteristics based on AC data. Among them, the operating condition characteristics are used to indicate the current operating condition of the energy storage system.

[0068] To illustrate, the collected DC data is preprocessed, such as detrending, removing DC components, and normalizing, to obtain operating condition characteristics. These operating condition characteristics can reflect whether the energy storage system is currently in a charging, discharging, or no-load state. Current fluctuations are inherently different under different modes. After adding operating condition characteristics, the arc detection model can specifically "calibrate" the arc characteristic threshold, avoiding mistaking normal charging and discharging fluctuations for arc signals.

[0069] The AC data is first windowed using a window function, and then its spectrum is obtained through a fast Fourier transform. Key frequency domain features, including fundamental amplitude, harmonic content, and total harmonic distortion, are extracted from the spectrum. These AC frequency domain features will show significant distortion when arc light occurs, thereby improving the accuracy of arc detection.

[0070] In the time domain of the AC waveform, AC time-domain features such as instantaneous current envelope, large signal pulse width, and current zero-crossing offset within half a cycle are extracted. These AC time-domain features can capture the transient pulse behavior during arc discharge, thus supplementing the deficiencies of frequency domain features and improving the accuracy of arc detection.

[0071] S103 inputs the operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics into the arc detection model to obtain the arc detection results.

[0072] Among them, the arc detection model is based on a convolutional neural network or a long short-term memory network (LSTM), or it can be a lightweight classifier based on a support vector machine (SVM), and so on.

[0073] The arc detection results include two possibilities: the presence of an arc in the current energy storage system and the absence of an arc. When the arc detection results indicate the presence of an arc, the arc extinguishing device can be quickly activated or the circuit can be cut off to ensure the safe operation of the energy storage system.

[0074] The arc detection method for energy storage systems provided in this application acquires current data from the DC and AC sides of the energy storage system, and then extracts operating condition features, AC frequency domain features, and AC time domain features from these data. Operating condition features reflect whether the energy storage system is currently charging, discharging, or unloaded. Current fluctuation patterns differ significantly under different operating conditions, and the arc detection model can dynamically adjust feature thresholds accordingly to reduce misjudgments caused by changes in operating conditions and improve the model's adaptability and robustness. Frequency domain features reveal phenomena such as changes in harmonic content and increased total harmonic distortion caused by arc occurrence, exhibiting good global identification capabilities. Time domain features capture transient pulse behaviors during arc discharge, such as current envelope fluctuations, pulse width changes, and zero-crossing offsets, thus effectively supplementing the frequency domain features. The combination of operating condition features, AC frequency domain features, and AC time domain features enhances the arc detection model's ability to identify different types of arcs (such as continuous arcs and intermittent arcs), improving detection accuracy.

[0075] In some embodiments of this application, such as Figure 2 As shown, the arc detection model includes a first branch, a second branch, and a third branch.

[0076] The operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics are input into the arc detection model to obtain the arc detection results, including:

[0077] The first feature is obtained by processing the operating condition characteristics and AC time domain characteristics through the first branch.

[0078] Schematic, the first branch is used to extract the operating status features and AC signal dynamic features of the energy storage system from the operating condition features and AC time domain features. Specifically, a one-dimensional convolutional neural network or a multilayer perceptron can be used to perform feature transformation on the statistical quantities such as the mean current, variance, and mutation rate in the operating condition features and the waveform morphology parameters and transient amplitude change rate in the AC time domain features, thereby generating a deep feature representation (i.e., the first feature) that can simultaneously reflect the operating states such as charging, discharging, and no-load and the AC signal change pattern.

[0079] The second feature is obtained by processing the AC frequency domain features through the second branch.

[0080] Schematic, the second branch can use a multi-layer convolutional structure or a spectrogram processing network to model frequency domain parameters such as fundamental amplitude, harmonic content, and total harmonic distortion, and extract the spectral distortion features caused by the electric arc, thereby forming a recognizable frequency domain feature (i.e., the second feature).

[0081] The third feature is obtained by processing the time-domain features of the communication through the third branch.

[0082] Schematic, the third branch can be based on temporal neural networks such as LSTM, GRU, or convolutional neural networks to extract transient pulse behaviors related to the electric arc in AC waveforms, thereby obtaining the third feature.

[0083] The arc detection result is determined based on the first, second, and third features. That is, the first, second, and third features are fused. Illustratively, features such as feature concatenation, attention weighting, or gating mechanisms can be used to integrate the first, second, and third features into a unified fused feature. This fused feature is then input into the fully connected layer and classification layer in the arc detection model to finally output the arc detection result.

[0084] Understandably, the processing of the first, second, and third branches is not sequential and can be performed simultaneously.

[0085] The arc detection method for energy storage systems provided in this application integrates the first, second, and third features, enabling the arc detection model to achieve information complementarity and contextual association under complex operating conditions. Furthermore, it maintains stable detection performance even when facing different modes and multiple disturbances (such as current amplitude fluctuations and environmental noise), significantly reducing the probability of false alarms and missed alarms.

[0086] In some embodiments of this application, the operating condition characteristics include DC signal characteristics and charge / discharge state. Determining the operating condition characteristics based on DC data includes:

[0087] The DC data is divided into multiple segments to obtain the characteristics of the DC signal, and the charging and discharging state is determined based on the direction indicated by the DC data.

[0088] To illustrate, because the DC current of an energy storage system varies widely, from 0A to hundreds of amperes, the operating state of the system differs significantly across different current ranges. Therefore, after acquiring the DC current signal, the sampling circuit board first divides the entire DC current variation range into several non-overlapping current intervals based on preset interval boundaries. Then, it compares the acquired real-time DC current value with the boundary values ​​of each interval to determine the target interval to which the DC current belongs. Finally, it maps the target interval to the corresponding DC signal characteristics for subsequent processing.

[0089] For example, when the rated DC current range of an energy storage system is 0A to 140A, it can be divided into three current ranges: 0–60A, 60–100A, and 100–140A. When the real-time current value is in the 0–60A range, it is mapped to a low-flow operating condition, and the corresponding three DC signal characteristics are represented as (1,0,0); when it is in the 60–100A range, it is mapped to a medium-flow operating condition, and the corresponding three DC signal characteristics are represented as (0,1,0); when it is in the 100–140A range, it is mapped to a high-flow operating condition, and the corresponding three DC signal characteristics are represented as (0,0,1).

[0090] Understandably, the amplitude differences between normal charging and discharging fluctuations and arc signals vary under different DC signal characteristics. By judging the characteristics of the DC signal, the arc detection model can use a more sensitive threshold under "low flow" conditions to capture weak arc signals; and under "high flow" conditions, it can increase the threshold to avoid misjudging strong normal fluctuations as arcs, thereby reducing the false alarm rate.

[0091] The first branch processes the operating condition characteristics and AC time-domain characteristics to obtain the first feature, including:

[0092] By reshaping the dimensions of multiple DC signal features, charging and discharging states, and AC time-domain features through the first branch, a first feature with higher expressive power is obtained.

[0093] In some embodiments of this application, the method further includes:

[0094] Get and disconnect status.

[0095] In some examples, under grid-connected mode, the energy storage system maintains frequency and phase synchronization with the public power grid through an inverter. The grid's equivalent impedance is relatively large, providing strong suppression of transient current changes. When the current suddenly changes at the beginning of an arc, the grid can quickly divert or absorb some of the pulse energy, relatively limiting the current spike amplitude and shortening its duration, resulting in a slight increase in harmonic content in the spectrum. At this time, the difference between the arc signal and normal harmonic interference decreases, requiring the arc detection model to improve sensitivity and narrow the threshold interval to accurately capture weak arc distortions.

[0096] In off-grid mode, the energy storage system independently regulates its output voltage and frequency. The inverter controller responds more directly to load changes, resulting in a smaller equivalent impedance and a significant increase in both the amplitude and duration of current pulses. When an arc occurs, transient pulses are more easily confused with current spikes caused by normal load switching. Simultaneously, factors such as the energy storage system's own oscillations and rectifier bridge conduction delays also contribute to high-frequency noise. In off-grid mode, the arc detection model places greater emphasis on time-domain pulse characteristics (such as envelope peak value and pulse width) and frequency band energy distribution to improve the ability to identify large-amplitude arc pulses.

[0097] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, and AC time-domain characteristics through the first branch, including:

[0098] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, grid connection and disconnection states, and AC time domain characteristics through the first branch.

[0099] In some embodiments of this application, the arc detection result includes a first score value arc and a second score value normal. The first score value is used to indicate the confidence level that an arc has occurred in the energy storage system, and the second score value is used to indicate the confidence level that no arc has occurred in the energy storage system.

[0100] After obtaining the arc detection results, the method also includes:

[0101] Based on the mapping relationship between the on-grid / off-grid status and the score increment, obtain the first score increment or the second score increment corresponding to the on-grid / off-grid status.

[0102] In this system, the first score increment corresponds to the first score value, and the second score value corresponds to the second score increment. The mapping relationship means that the score increment values ​​differ for different grid-connected and off-grid states. For example, the first score increment value for the grid-connected state is different from the first score increment value for the off-grid state. Similarly, the second score increment value for the grid-connected state is different from the second score increment value for the off-grid state. In some examples, because the arc signal is weak and easily missed when grid-connected, and the normal fluctuation amplitude is large and easily false alarms when off-grid, the second score increment value for the grid-connected state is a small increment. Fine-tuning this small increment can significantly reduce the misjudgment of normal harmonic distortion as an arc. The second score increment value for the off-grid state is a large increment. A larger score increment can effectively raise the "no arc" confidence threshold, eliminating large pulse false alarms at the source.

[0103] After determining the offline status, the specific value of the first score increment or the second score increment can be determined based on the aforementioned mapping relationship.

[0104] The first score is adjusted based on the first score increment; or the second score is adjusted based on the second score increment. That is, the sum of the first score increment and the first score is used as the adjusted first score, and the sum of the second score increment and the second score is used as the adjusted second score.

[0105] It should be noted that only one of the first rating value and the second rating value needs to be adjusted. For example, subtracting the first rating increment from the first rating value and adding the second rating increment to the second rating value both increase the probability of normal. Therefore, only one of the two needs to be adjusted.

[0106] The first score value is compared with the adjusted second score value, or the adjusted first score value is compared with the second score value, and the final arc detection result is determined based on the comparison result.

[0107] Schematic, a first score value is compared with an adjusted second score value. If the first score value is greater than the adjusted second score value, the arc detection result is determined to be an arc anomaly; if the first score value is less than the adjusted second score value, the arc detection result is determined to be a normal arc. In other examples, the adjusted first score value can also be compared with the second score value when the first score value is adjusted.

[0108] The arc detection method for energy storage systems provided in this application embodiment can adjust the first score value of the original output of the arc detection model by the first score increment, or adjust the second score value of the original output of the arc detection model by the second score increment, thereby improving the false alarm suppression capability in a targeted manner under different grid connection and off-grid modes.

[0109] In some embodiments of this application, the AC data includes an AC time-domain signal, and determining the AC time-domain characteristics based on the AC data includes:

[0110] The number of sampling points with current values ​​greater than the current threshold and the number of sampling points with current values ​​lower than the current threshold in the AC time-domain signal are counted separately to obtain the first time-domain statistical feature and the second time-domain statistical feature. The AC time-domain feature includes the AC time-domain signal, the first time-domain statistical feature and the second time-domain statistical feature.

[0111] During the arc initiation phase of energy storage arcing, the AC time-domain signal often exhibits sudden high-amplitude spikes or brief suppression. Based on this, in some examples, threshold statistical processing is performed on the acquired AC time-domain signal. For a certain preset current threshold, the number of points where the current exceeds the threshold and the number of points where the current is below the threshold are counted among all sampling points within a window, thereby obtaining the first time-domain statistical feature and the second time-domain statistical feature.

[0112] The first branch processes the operating condition characteristics and AC time-domain characteristics to obtain the first feature, including:

[0113] The first feature is obtained by processing the operating condition characteristics, the first time-domain statistical characteristics, and the second time-domain statistical characteristics through the first branch.

[0114] The third branch processes the time-domain features of the AC signal to obtain the third feature, which includes:

[0115] The third feature is obtained by processing the AC time-domain signal through the third branch.

[0116] The arc detection method for energy storage systems provided in this application addresses the issue that arcs often involve extremely short-duration, high-amplitude current spikes or rapid decay during the arc initiation stage. These transients occupy a very short time and have relatively low energy within the overall time-domain signal, insufficient to significantly alter the signal's mean or variance. This application obtains a first time-domain statistical feature by counting the number of sampling points exceeding a threshold. Within the first few milliseconds after the spike appears, the accumulated number of "abnormal points" is immediately quantified using the first time-domain statistical feature, thus providing the earliest warning information. Furthermore, in some arcing types, the current waveform may briefly decrease or exhibit empty flame (the arc is not conducting but high voltage is present). Traditional peak detection struggles to distinguish this type of weak suppression. Counting the number of points below a threshold (i.e., the second time-domain statistical feature) can reflect this transient current drop. By accumulating the number of "low-amplitude" sampling points, signal anomalies can also be captured before the arc stabilizes. Therefore, the first and second time-domain statistical features can identify arcs in a short time.

[0117] In some embodiments of this application, both the second branch and the third branch include multiple convolutional blocks. The number of convolutional blocks in the second branch and the third branch may be the same or different, and the network layers contained in each convolutional block may be the same or different.

[0118] The second branch processes the AC frequency domain features to obtain the second feature, including:

[0119] The second feature is obtained by convolving the AC frequency domain features with multiple convolutional blocks in the second branch.

[0120] Schematic, AC frequency domain features are sequentially input into multiple connected convolutional blocks. These blocks extract higher-order harmonics and broadband noise features from the AC frequency domain features. Each convolutional block expands the receptive field gradually through downsampling or stride control, forming a second feature. By learning multi-scale distortion features in the AC frequency domain through multiple convolutional blocks, the arc detection model can identify both weak harmonic drift and capture the broadband noise boost caused by the arc.

[0121] The third branch processes the time-domain features of the AC signal to obtain the third feature, which includes:

[0122] The third feature is obtained by convolving the temporal features of the communication through multiple convolutional blocks in the third branch.

[0123] Indicatively, a high-dimensional representation (i.e., the third feature) is extracted from the AC time-domain features through multiple sequentially connected convolutional blocks, providing a rich and robust time-domain representation for the arc detection model.

[0124] In some examples, such as Figure 2 As shown, both the second and third branches consist of four convolutional blocks connected in sequence, plus a dimension reshaping layer. According to the connection order, the network layers within each of the four convolutional blocks are as follows: the first convolutional block includes a normalization layer, a convolutional layer, and a ReLU layer; the second convolutional block includes a pooling layer, a convolutional layer, and a ReLU layer; the third convolutional block includes a convolutional layer and a ReLU layer; and the fourth convolutional block includes a convolutional layer and a ReLU layer.

[0125] In some embodiments of this application, the arc detection model is obtained in the following manner:

[0126] Obtain current sample data and corresponding arc label information.

[0127] The pre-built arc detection model is trained based on current sample data and arc label information to obtain an initial arc detection model.

[0128] The initial arc detection model is fine-tuned using both on-grid and off-grid data to obtain the final arc detection model.

[0129] like Figure 3 As shown, current sample data is acquired, and operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics are extracted from the current sample data. The operating condition characteristics include DC signal characteristics, charging / discharging status, and grid-connected / off-grid status. These operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics are input into a pre-built arc detection model, and the model is trained using arc label information to obtain an initial arc detection model. Based on the test results of the initial arc detection model, for cases of false alarms and missed alarms, a small amount of grid-connected / off-grid data is used to fine-tune the initial arc detection model. Through fine-tuning training, the parameters of the convolutional network and fully connected layers of the initial arc detection model are adjusted to improve the model's accuracy.

[0130] Finally, the first score value output by the final arc detection model is adjusted by the first score increment mentioned above, or the second score value output by the final arc detection model is adjusted by the second score increment. The final arc detection result is determined based on the first score value and the adjusted second score value, or the adjusted first score value and the adjusted second score value, or the adjusted first score value and the adjusted second score value.

[0131] like Figure 4 As shown in the figure, this application embodiment also provides an arc detection device for an energy storage system, which includes a data acquisition module 401, a feature determination module 402, and a detection module 403.

[0132] Data acquisition module 401 is used to acquire current data of the energy storage system, including DC data and AC data;

[0133] The feature determination module 402 is used to determine the operating condition features based on DC data, and to determine the AC frequency domain features and AC time domain features based on AC data. The operating condition features are used to indicate the current operating condition of the energy storage system.

[0134] The detection module 403 is used to input the operating condition characteristics, AC frequency domain characteristics and AC time domain characteristics into the arc detection model to obtain the arc detection results.

[0135] In some embodiments of this application, the arc detection model includes a first branch, a second branch, and a third branch;

[0136] The detection module includes:

[0137] The first unit is used to process the operating condition characteristics and AC time domain characteristics through the first branch to obtain the first feature;

[0138] The second unit is used to process the AC frequency domain characteristics through the second branch to obtain the second characteristic;

[0139] The third unit is used to process the AC time-domain features through the third branch to obtain the third feature;

[0140] The fusion unit is used to determine the arc detection result based on the first feature, the second feature, and the third feature.

[0141] In some embodiments of this application, the operating condition characteristics include DC signal characteristics and charging / discharging state, and the characteristic determination module includes:

[0142] The operating condition characteristic determination unit is used to perform binning operations on DC data to obtain the characteristics of multiple DC signal segments; and to determine the charging and discharging state based on the direction indicated by the DC data.

[0143] The first unit is specifically used for:

[0144] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, and AC time-domain characteristics through the first branch.

[0145] In some embodiments of this application, the arc detection device of the energy storage system further includes:

[0146] The status acquisition module is used to acquire and disconnection status.

[0147] The first unit is specifically used for:

[0148] The first feature is obtained by reshaping the dimensions of multiple DC signal characteristics, charging and discharging states, grid connection and disconnection states, and AC time domain characteristics through the first branch.

[0149] In some embodiments of this application, the arc detection result includes a first score value and a second score value. The first score value is used to indicate the confidence level that an arc has occurred in the energy storage system, and the second score value is used to indicate the confidence level that no arc has occurred in the energy storage system.

[0150] The arc detection device of the energy storage system also includes a result adjustment module for:

[0151] Based on the mapping relationship between the grid connection and disconnection status and the score increment, obtain the first score increment or the second score increment corresponding to the grid connection and disconnection status;

[0152] The first score value is adjusted based on the first score increment; or the second score value is adjusted based on the second score increment.

[0153] The first score is compared with the adjusted second score, or the adjusted first score is compared with the second score, and the final arc detection result is determined based on the comparison result.

[0154] In some embodiments of this application, the AC data includes an AC time-domain signal, and determining the AC time-domain characteristics based on the AC data includes:

[0155] The number of sampling points with current values ​​greater than the current threshold and the number of sampling points with current values ​​lower than the current threshold in the AC time-domain signal are counted respectively to obtain the first time-domain statistical feature and the second time-domain statistical feature. The AC time-domain feature includes the AC time-domain signal, the first time-domain statistical feature and the second time-domain statistical feature.

[0156] The first branch processes the operating condition characteristics and AC time-domain characteristics to obtain the first feature, including:

[0157] The first feature is obtained by processing the operating condition characteristics, the first time-domain statistical characteristics, and the second time-domain statistical characteristics through the first branch;

[0158] The third branch processes the time-domain features of the AC signal to obtain the third feature, which includes:

[0159] The third feature is obtained by processing the AC time-domain signal through the third branch.

[0160] In some embodiments of this application, both the second branch and the third branch include multiple convolutional blocks;

[0161] The second unit is specifically used for:

[0162] The AC frequency domain features are convolved through multiple convolutional blocks in the second branch to obtain the second feature;

[0163] The third unit is specifically used for:

[0164] The third feature is obtained by convolving the temporal features of the communication through multiple convolutional blocks in the third branch.

[0165] In some embodiments of this application, the arc detection model is obtained in the following manner:

[0166] Acquire current sample data and corresponding arc label information;

[0167] The pre-built arc detection model is trained based on current sample data and arc label information to obtain an initial arc detection model.

[0168] The initial arc detection model is fine-tuned using both on-grid and off-grid data to obtain the final arc detection model.

[0169] The arc detection device for an energy storage system provided in this application acquires current data from the DC and AC sides of the energy storage system, and then extracts operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics from them. Operating condition characteristics reflect whether the energy storage system is currently charging, discharging, or unloaded. Current fluctuation patterns differ significantly under different operating conditions, and the arc detection model can dynamically adjust feature thresholds accordingly to reduce misjudgments caused by changes in operating conditions and improve the model's adaptability and robustness. Frequency domain characteristics reveal phenomena such as changes in harmonic content and increased total harmonic distortion caused by arc occurrence, exhibiting good global identification capabilities. Time domain characteristics capture transient pulse behaviors during arc discharge, such as current envelope fluctuations, pulse width changes, and zero-crossing offsets, thus effectively supplementing the frequency domain characteristics. The combination of operating condition characteristics, AC frequency domain characteristics, and AC time domain characteristics enhances the arc detection model's ability to identify different types of arcs (such as continuous arcs and intermittent arcs), improving detection accuracy.

[0170] Based on any of the above embodiments, another embodiment of this application also provides an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logic instructions in the memory 530 to execute the arc detection method of the energy storage system described above.

[0171] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0172] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0173] On the other hand, embodiments of this application also provide a storage medium storing a plurality of instructions adapted for loading by a processor to execute the arc detection method of the energy storage system provided in the above embodiments.

[0174] On the other hand, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described arc detection method for the energy storage system.

[0175] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0176] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0177] The above provides a detailed description of the arc detection method, apparatus, equipment, and medium for an energy storage system provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting electric arc in an energy storage system, characterized in that, The method includes: Acquire current data from the energy storage system, the current data including DC data and AC data; Operating condition characteristics are determined based on the DC data, and AC frequency domain characteristics and AC time domain characteristics are determined based on the AC data. The operating condition characteristics are used to indicate the current operating condition of the energy storage system. The operating condition characteristics, the AC frequency domain characteristics, and the AC time domain characteristics are input into the arc detection model to obtain the arc detection results.

2. The arc detection method for an energy storage system according to claim 1, characterized in that, The arc detection model includes a first branch, a second branch, and a third branch; The step of inputting the operating condition characteristics, the AC frequency domain characteristics, and the AC time domain characteristics into the arc detection model to obtain the arc detection result includes: The first feature is obtained by processing the operating condition features and the AC time domain features through the first branch; The second branch processes the AC frequency domain features to obtain the second feature; The third branch is used to process the AC time-domain features to obtain the third feature; The arc detection result is determined based on the first feature, the second feature, and the third feature.

3. The arc detection method for an energy storage system according to claim 2, characterized in that, The operating condition characteristics include DC signal characteristics and charging / discharging status. Determining the operating condition characteristics based on the DC data includes: The DC data is divided into multiple segments to obtain DC signal characteristics; and the charging and discharging state is determined based on the direction indicated by the DC data. The step of processing the operating condition features and the AC time-domain features through the first branch to obtain the first feature includes: The first feature is obtained by reshaping the dimensions of the multi-segment DC signal features, the charging and discharging states, and the AC time-domain features through the first branch.

4. The arc detection method for an energy storage system according to claim 3, characterized in that, The method further includes: Get and log off-network status; The step of reshaping the dimensions of the multiple DC signal features, the charging and discharging states, and the AC time-domain features through the first branch to obtain the first feature includes: The first feature is obtained by reshaping the dimensions of the multi-segment DC signal features, the charging and discharging state, the grid connection and disconnection state, and the AC time domain features through the first branch.

5. The arc detection method for an energy storage system according to claim 4, characterized in that, The arc detection result includes a first score and a second score. The first score is used to indicate the confidence level that the energy storage system has generated an arc, and the second score is used to indicate the confidence level that the energy storage system has not generated an arc. After obtaining the arc detection result, the method further includes: Based on the mapping relationship between the on-grid and off-grid states and the score increment, obtain the first score increment or the second score increment corresponding to the on-grid and off-grid states; The first score value is adjusted based on the first score increment; or the second score value is adjusted based on the second score increment. The first score value is compared with the adjusted second score value, or the adjusted first score value is compared with the second score value, and the final arc detection result is determined based on the comparison result.

6. The arc detection method for an energy storage system according to claim 2, characterized in that, The AC data includes an AC time-domain signal, and determining the AC time-domain characteristics based on the AC data includes: The number of sampling points with current greater than the current threshold and the number of sampling points with current lower than the current threshold in the AC time-domain signal are counted respectively to obtain the first time-domain statistical feature and the second time-domain statistical feature. The AC time-domain feature includes the AC time-domain signal, the first time-domain statistical feature and the second time-domain statistical feature. The step of processing the operating condition features and the AC time-domain features through the first branch to obtain the first feature includes: The first feature is obtained by processing the operating condition feature, the first time-domain statistical feature, and the second time-domain statistical feature through the first branch; The process of processing the AC time-domain features through the third branch to obtain the third feature includes: The third feature is obtained by processing the AC time-domain signal through the third branch.

7. The arc detection method for an energy storage system according to claim 2, characterized in that, Both the second branch and the third branch include multiple convolutional blocks; The step of processing the AC frequency domain features through the second branch to obtain the second feature includes: The AC frequency domain features are convolved by multiple convolutional blocks in the second branch to obtain the second feature; The process of processing the AC time-domain features through the third branch to obtain the third feature includes: The third feature is obtained by convolving the AC time-domain features with multiple convolutional blocks in the third branch.

8. The arc detection method for an energy storage system according to any one of claims 1 to 7, characterized in that, The arc detection model was obtained in the following manner: Acquire current sample data and corresponding arc label information; The pre-built arc detection model is trained based on the current sample data and the arc label information to obtain an initial arc detection model. The initial arc detection model is fine-tuned using both on-grid and off-grid data to obtain the final arc detection model.

9. An arc detection device for an energy storage system, characterized in that, The device includes: The data acquisition module is used to acquire the current data of the energy storage system, the current data including DC data and AC data; The feature determination module is used to determine the operating condition features based on the DC data, and to determine the AC frequency domain features and AC time domain features based on the AC data. The operating condition features are used to indicate the current operating condition of the energy storage system. The detection module is used to input the operating condition characteristics, the AC frequency domain characteristics, and the AC time domain characteristics into the arc detection model to obtain the arc detection results.

10. An electronic device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor runs the computer program in the memory to perform the operation in the arc detection method of the energy storage system according to any one of claims 1 to 8.

11. A storage medium, characterized in that, The storage medium stores a plurality of instructions adapted for loading by a processor to execute the steps of the arc detection method of the energy storage system according to any one of claims 1 to 8.

Images