Charging cabinet operation state monitoring method, device and system based on Internet of Things data

Abstract

The invention provides a charging cabinet operation state monitoring method, device and system based on Internet of Things data, and relates to the technical field of artificial intelligence, data normalization is carried out based on absolute safety boundary in combination with local statistical characteristics, extreme abnormal data can be quickly filtered, data characteristics of each channel can be adapted, and the reliability of the charging cabinet operation state monitoring is improved. And the core time sequence characteristics are reserved to the maximum extent while the dimension difference is eliminated. And further constructing a channel interaction fusion vector, and extracting boundary violation sensitive features to execute state monitoring. A channel interaction fusion vector can capture a composite abnormal mode corresponding to the data, and the comprehensiveness of abnormal recognition is improved; and the boundary violation sensitive features can accurately capture the early-stage precursor of the fault, so that the conversion from passive response to active early warning is realized. In conclusion, the identification limitation in the prior art can be broken through, the operation state of the charging cabinet can be accurately identified from massive high-dimensional time sequence data, the operation safety early warning capability can be enhanced, and the intelligent and large-scale level of operation and maintenance management can be improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, device and system for monitoring the operating status of charging cabinets based on Internet of Things (IoT) data. Background Technology

[0002] With the rapid popularization of electric vehicles, electric bicycles, and various portable electronic devices, charging cabinets, as centralized and shared power supply infrastructure, are being deployed on an ever-increasing scale. The operational safety and reliability of charging cabinets are directly related to user property, electrical safety, and even public safety. Abnormal conditions such as overvoltage or overcurrent can lead to serious consequences such as equipment damage and fires. Therefore, real-time and accurate online monitoring and early fault warning of charging cabinet operation status has become an urgent and critical technical requirement in operation and maintenance management.

[0003] Currently, IoT-based monitoring technology provides the data foundation for this need. By deploying a sensor network inside the charging cabinet, time-series data of key physical quantities such as current, voltage, and power during the charging process can be collected in real time. However, how to automatically and accurately identify various normal and abnormal operating states from this massive amount of high-dimensional time-series data still faces a series of technical challenges. Summary of the Invention

[0004] The purpose of this invention is to provide a method, device and system for monitoring the operating status of charging cabinets based on Internet of Things data, which can overcome the identification limitations of existing technologies and automatically and accurately identify various normal and abnormal operating states from massive, high-dimensional time-series data.

[0005] In a first aspect, the present invention provides a method for monitoring the operational status of a charging cabinet based on Internet of Things (IoT) data. The method includes: determining operational monitoring data of the target charging cabinet through its IoT sensor network; normalizing the operational monitoring data based on a preset absolute safety boundary and the local statistical features corresponding to the monitoring sequences of each monitoring channel in the operational monitoring data to generate target monitoring data; constructing a channel interaction fusion vector for the target monitoring data and determining the boundary violation sensitive features of the target monitoring data; and classifying and predicting the target monitoring data based on the channel interaction fusion vector and the boundary violation sensitive features to determine the operational status classification result of the target charging cabinet.

[0006] In conjunction with the first aspect, this embodiment of the invention provides a first implementation of the first aspect, wherein the step of normalizing the operational monitoring data based on a preset absolute safety boundary and the local statistical features corresponding to the monitoring sequences of each monitoring channel in the operational monitoring data to generate target monitoring data includes: determining the local average level corresponding to the monitoring sequences of each monitoring channel in the operational monitoring data based on a preset local window length; determining the dynamic normalization boundary corresponding to the local average level based on the preset absolute safety boundary; and normalizing the monitoring sequences of each monitoring channel based on the dynamic normalization boundary to construct the target monitoring data corresponding to the operational monitoring data.

[0007] In conjunction with the first aspect, this embodiment of the invention provides a second implementation of the first aspect, wherein the step of constructing the channel interaction fusion vector of the target monitoring data includes: using a preset coupling kernel tensor to perform inter-channel information fusion on the monitoring sequence of each monitoring channel of the target monitoring data to generate an intermediate tensor corresponding to the target monitoring data; performing singular value decomposition on the intermediate tensor to determine a first change pattern of the target monitoring data for the joint time dimension and a second change pattern for the data depth dimension; and performing data filtering processing on the target monitoring data based on the first singular vector matrix corresponding to the first change pattern and the second singular vector matrix corresponding to the second change pattern to generate the channel interaction fusion vector of the target monitoring data.

[0008] In conjunction with the first aspect, this embodiment of the invention provides a third implementation of the first aspect, wherein the step of determining the boundary violation sensitive features of the target monitoring data includes: determining the boundary violation event data corresponding to the target monitoring data based on preset boundary violation parameters; determining the boundary proximity data corresponding to the target monitoring data based on the normalized relative boundary corresponding to the normalization processing; and determining the boundary violation sensitive features of the target monitoring data based on the boundary violation event data and the boundary proximity data.

[0009] In conjunction with the first aspect, this embodiment of the invention provides a fourth implementation of the first aspect, wherein the step of classifying and predicting the target monitoring data based on the channel interaction fusion vector and boundary violation sensitive features to determine the operation status classification result of the target charging cabinet includes: fusing the boundary violation sensitive features into the channel interaction fusion vector to construct a focused feature tensor corresponding to the target monitoring data; and using a pre-built operation status monitoring model to classify and predict the focused feature tensor to determine the operation status classification result corresponding to the target charging cabinet.

[0010] In conjunction with the first aspect, this embodiment of the invention provides a fifth implementation of the first aspect, wherein the step of fusing boundary violation sensitive features into the channel interaction fusion vector to construct a focused feature tensor corresponding to the target monitoring data includes: using a preset gating fusion mechanism to fuse boundary violation sensitive features into the channel interaction fusion vector to construct an initial fusion feature tensor; and adaptively weighting the initial fusion feature tensor based on preset multi-dimensional attention weights to construct a focused feature tensor.

[0011] In conjunction with the first aspect, this embodiment of the invention provides a sixth implementation of the first aspect, wherein the step of using a pre-built operation status monitoring model to classify and predict the focused feature tensor and determine the operation status classification result corresponding to the target charging cabinet includes: using a multi-prototype metric learning mechanism to determine the probability distribution prediction result of each monitoring channel of the focused feature tensor based on the corresponding pattern summary vector; and determining the operation status classification result corresponding to the target charging cabinet based on the probability distribution prediction result.

[0012] In conjunction with the first aspect, this embodiment of the invention provides a seventh implementation of the first aspect, wherein the operation status monitoring model is constructed based on the channel interaction fusion vector and boundary violation sensitive features of the preset training samples; the loss function of the operation status monitoring model includes a violation-aware weighted classification loss and a coupled-mode filtering consistency regularization loss; wherein the violation-aware weighted classification loss is calculated based on the dynamic weights of the training samples, and the dynamic weights are determined based on the boundary violation sensitive features; the coupled-mode filtering consistency regularization loss is used to constrain the consistency of the filtering parameters corresponding to the channel interaction fusion vector.

[0013] Secondly, embodiments of the present invention provide a charging cabinet operation status monitoring device based on Internet of Things (IoT) data. The device includes: a data monitoring module for determining operation monitoring data of the target charging cabinet through an IoT sensor network; a data preprocessing module for normalizing the operation monitoring data based on a preset absolute safety boundary and the local statistical features corresponding to the monitoring sequences of each monitoring channel in the operation monitoring data, generating target monitoring data; a data processing module for constructing a channel interaction fusion vector of the target monitoring data and determining the boundary violation sensitive features of the target monitoring data; and an execution module for classifying and predicting the target monitoring data based on the channel interaction fusion vector and the boundary violation sensitive features, determining the operation status classification result of the target charging cabinet.

[0014] Thirdly, embodiments of the present invention provide a charging cabinet operation status monitoring system based on Internet of Things data. The system is equipped with the apparatus described in the above embodiments and is used to execute the methods of any of the above embodiments.

[0015] The embodiments of this invention bring the following beneficial effects: This invention provides a method, device, and system for monitoring the operational status of charging cabinets based on IoT data. By using absolute safety boundaries, it can quickly filter extreme abnormal data, avoiding interference from such data at the source. Furthermore, by adapting local statistical features to the characteristics of data from each channel, it eliminates dimensional differences while maximizing the preservation of core temporal characteristics. To this end, the construction of channel interaction fusion vectors can capture complex abnormal patterns that cannot be reflected by a single channel, improving the comprehensiveness of anomaly identification. Boundary violation sensitive features can break the limitation of post-event judgment of "exceeding the standard equals an alarm," accurately capturing early warning signals of faults, allowing sufficient time for operation and maintenance, and shifting from passive response to proactive early warning. In summary, the embodiments of this invention can effectively overcome the technical challenges of automatically and accurately identifying the operational status of charging cabinets under massive high-dimensional time-series data in the prior art. This not only strengthens the early warning capability for charging cabinet operational safety but also improves the intelligence and scalability of operation and maintenance management.

[0016] Other features and advantages of the invention will be set forth in the following description, or some features and advantages may be inferred from the description or determined without doubt, or may be learned by practicing the techniques described above.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a method for monitoring the operating status of a charging cabinet based on Internet of Things (IoT) data, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another method for monitoring the operating status of a charging cabinet based on Internet of Things (IoT) data, provided in an embodiment of the present invention; Figure 3 A schematic diagram illustrating the working principle of a boundary violation sensitive feature and its response to abnormal states, provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a classification prediction process provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a charging cabinet operation status monitoring device based on Internet of Things data, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] This invention provides a method, device, and system for monitoring the operating status of charging cabinets based on Internet of Things (IoT) data. It can overcome the identification limitations of existing technologies and automatically and accurately identify various normal and abnormal operating states from massive, high-dimensional time-series data.

[0023] To facilitate understanding, the method for monitoring the operating status of a charging cabinet based on Internet of Things (IoT) data, as provided in this embodiment of the invention, will be described first. (Refer to...) Figure 1 The method includes the following steps: Step S102: Determine the operation monitoring data of the target charging cabinet through the IoT sensor network of the target charging cabinet.

[0024] The target charging cabinet is the specific monitoring entity required for real-time monitoring of its operational status and fault early warning in this embodiment of the invention. It can function as an independent unit for refined monitoring of a single charging cabinet, or as a basic monitoring unit in a batch monitoring scenario. An IoT sensor network is pre-deployed at key operating nodes inside the target charging cabinet, including but not limited to charging modules, circuit interfaces, and heat dissipation systems. This network serves as the core carrier for data acquisition, relying on IoT communication technology to collect various status data during the operation of the target charging cabinet in real time and transmit the collected raw data to the back-end processing unit in real time. This IoT sensor network is a distributed sensing network, which can consist of multiple different types of sensors to specifically collect multi-dimensional physical quantity data such as current, voltage, power, temperature, and humidity, which are strongly correlated with the operational safety of the charging cabinet.

[0025] The corresponding operational monitoring data consists of raw time-series data collected through the aforementioned IoT sensor network, accurately reflecting the real-time status of the target charging cabinet throughout its operation. This data may include multi-dimensional physical quantities strongly correlated with the charging cabinet's operational safety, such as current, voltage, and power, exhibiting high-dimensional and time-series characteristics.

[0026] Step S104: Based on the preset absolute safety boundary and the local statistical features corresponding to the monitoring sequence of each monitoring channel in the operation monitoring data, the operation monitoring data is normalized to generate target monitoring data.

[0027] The preset absolute safety boundary refers to the pre-set safety threshold range of each monitored physical quantity based on the charging cabinet equipment parameters, industry safety standards, and engineering practice experience. It serves as the underlying constraint for determining whether data is in an extremely dangerous state. Data streams exceeding this boundary can be directly identified as severely abnormal and prioritized for processing. In the aforementioned sensor network, the data acquisition link of a single sensor or a group of similar sensors constitutes a monitoring channel. Its monitoring sequence is a set of time-series data collected by a single monitoring channel over a continuous time dimension, reflecting the dynamic change trend of the corresponding physical quantity during the charging process (e.g., a sequence of current values ​​collected per second within one minute by a current channel). Local statistical features are statistical indicators calculated for the monitoring sequence of a single monitoring channel within a preset time window (e.g., 10 seconds, 30 seconds). Core metrics include mean, variance, peak value, trough value, and trend slope, reflecting the local fluctuation patterns and distribution characteristics of the time-series data and avoiding accidental interference from data at a single time point.

[0028] In summary, this invention, based on a preset absolute safety boundary as a rigid underlying constraint, can quickly identify and filter extreme anomaly data exceeding the safety threshold, thus avoiding interference from extreme data on subsequent feature extraction and classification prediction from the source. Simultaneously, it performs differentiated processing based on the independent local statistical characteristics of each monitoring channel, rather than using uniform standardization rules. This allows the normalization process to accurately adapt to the unique fluctuation patterns and distribution characteristics of different physical quantity time-series data, maximizing the preservation of core time-series features related to the charging cabinet's operating status in the original data. In conclusion, this approach eliminates the problems of dimensional differences and uneven distribution of data from different monitoring channels, mapping high-dimensional heterogeneous time-series data to a unified numerical range. It also filters out the accidental interference from data at a single time point, making the generated target monitoring data more closely reflect the actual operating conditions of the charging cabinet.

[0029] Step S106: Construct the channel interaction fusion vector of the target monitoring data and determine the boundary violation sensitive features of the target monitoring data.

[0030] The channel interaction fusion vector is a high-dimensional feature vector formed by cross-dimensional correlation and integration of target monitoring data from multiple monitoring channels (e.g., constructed through feature fusion algorithms; in one implementation, attention mechanisms, convolutional fusion, matrix splicing, etc. can be used). It can capture the collaborative change relationship between data from different channels (e.g., the correlation between abnormal fluctuations in the current channel and sudden changes in the voltage channel).

[0031] Boundary violation sensitive features are the critical state features in the target monitoring data that are close to or slightly touch the preset absolute safety boundary. These features include the fluctuation amplitude, duration, and rate of change of the critical data, as well as the collaborative occurrence pattern of critical states in multiple channels. This is to capture early warning signals before a fault occurs and avoid the limitation of post-event judgment of "exceeding the standard and then alarming".

[0032] Step S108: Based on the channel interaction fusion vector and boundary violation sensitive features, classify and predict the target monitoring data to determine the classification result of the target charging cabinet's operating status.

[0033] Classification prediction based on channel interaction fusion vectors can fully utilize the captured collaborative correlation patterns of multiple monitoring channels to accurately identify complex anomaly patterns that cannot be reflected by single channel features, avoiding the missed detection problems caused by traditional single-parameter monitoring and improving the comprehensiveness of anomaly state identification. Classification prediction based on boundary violation sensitive features can accurately anchor critical anomaly signals close to the absolute safety boundary, effectively capturing early precursors of faults, breaking the limitation of "exceeding the standard and triggering an alarm" in post-event judgment, and allowing sufficient time for operation and maintenance. In summary, by integrating the complementary information of the two types of features, the operating status corresponding to the target monitoring data can be accurately classified. It can clearly distinguish basic states such as normal operation, minor anomalies, and severe anomalies, and further locate specific anomaly types such as overvoltage and overcurrent, improving the accuracy and directionality of status judgment. At the same time, the embodiments of this invention are fully automated, which can efficiently process the monitoring data of batch charging cabinets, adapt to large-scale cluster operation and maintenance scenarios, significantly reduce manual operation and maintenance costs, and improve operation and maintenance efficiency.

[0034] In summary, this invention can quickly filter extreme abnormal data through absolute safety boundaries, avoiding interference from such data at the source. Furthermore, by adapting local statistical features to the characteristics of each channel's data, it eliminates dimensional differences while preserving core temporal features to the maximum extent. To this end, the construction of channel interaction fusion vectors can capture complex abnormal patterns that cannot be reflected by a single channel, improving the comprehensiveness of anomaly identification. Boundary violation sensitive features can overcome the limitations of post-event judgment ("exceeding the standard triggers an alarm"), accurately capturing early warning signals of faults and allowing sufficient time for maintenance and repair, shifting from passive response to proactive early warning. In conclusion, this invention effectively addresses the technical challenges of automatically and accurately identifying the operating status of charging cabinets under massive high-dimensional time-series data in existing technologies. It strengthens the early warning capability for charging cabinet operation safety and improves the intelligence and scalability of operation and maintenance management.

[0035] Based on the above embodiments, this invention also provides another method for monitoring the operating status of charging cabinets based on Internet of Things (IoT) data, referring to... Figure 2 The method includes the following steps: Step S202: Determine the operation monitoring data of the target charging cabinet through the IoT sensor network of the target charging cabinet.

[0036] Based on the above embodiments, the corresponding data can be synchronously collected at a fixed sampling frequency through an IoT sensor network deployed inside the charging cabinet, capturing key physical quantity time-series data during the charging process. In this embodiment, the data mainly includes current, voltage, and power sequences. Current, voltage, and power sequences are three core data types in the charging cabinet's operational status monitoring, directly reflecting the electrical safety and energy interaction nature of the charging process. The current sequence characterizes load changes and overcurrent risks, the voltage sequence indicates power supply stability and insulation anomalies, and the power sequence comprehensively reflects energy conversion efficiency and abnormal power consumption patterns. These three types of data together constitute the basis of multi-dimensional time-series characteristics. Furthermore, by determining their coupling relationship, composite anomaly patterns can be captured. Combining their respective absolute safety boundaries and local statistical characteristics, accurate identification of early fault precursors and efficient discrimination of collaborative anomalies can be achieved, significantly improving the comprehensiveness and early warning capabilities of status monitoring.

[0037] The monitoring data for each charging process constitutes a raw data sample, which contains the complete trajectory of the three physical quantities changing over time from the start to the end of charging or within a specific monitoring period.

[0038] Correspondingly, the original data is defined as multi-channel time series data, and the original data sample is set as follows: For it, which includes three synchronous monitoring sequences of current, voltage, and power, the corresponding dimensions are: The data is collected synchronously by sensors in the charging cabinet. Each row represents a sequence of physical quantities changing over time. This indicates the length of the time series, which is the total length of the time series for a single sample. Represents the original data sample The 0th row corresponds to the entire current monitoring timing sequence; Represents the original data sample The first row corresponds to the entire voltage monitoring timing sequence; Represents the original data sample The second line corresponds to the entire power monitoring timing sequence.

[0039] Furthermore, data preprocessing (normalization) is performed on the corresponding data. The current, voltage, and power monitoring sequence data of the charging cabinet have clearly defined physical safety operating limits. Actual operating data fluctuates within the normal range, while abnormal states often manifest as slight boundary violations or drastic changes near the boundaries. Conventional global max-min normalization blurs the meaning of physical boundaries and is overly sensitive to abrupt changes in the sequence, leading to excessive compression of normal fluctuations after normalization. Z-score standardization fails to reflect the absolute safety range of the data. These shortcomings weaken the model's ability to identify boundary violations and abnormal fluctuations.

[0040] Regarding step S104 above, this invention employs a dynamic dual-boundary normalization method. This method not only references the prior physical boundary (i.e., the absolute safety boundary) but also dynamically calculates the normalized boundary at each time step by incorporating local statistical characteristics. Furthermore, by extending the normalized value range beyond the baseline interval, it retains and even highlights out-of-bounds information, thereby enhancing the model's sensitivity to behavior and abnormal states near the boundary. In specific implementation, refer to steps S204-S208 below.

[0041] Step S204: Based on the preset local window length, determine the local average level corresponding to the monitoring sequence of each monitoring channel in the running monitoring data.

[0042] Local average levels can characterize the operational data benchmark of a corresponding monitoring channel at different time periods, reflecting the local stationarity characteristics of time-series data. Specifically, the local average level can be calculated moment-by-moment using a sliding time window mechanism, such as by presetting the local window length. This is used to define the statistical range of local data. Its value can be adaptively adjusted according to the operating conditions of the charging cabinet and the required monitoring accuracy. In one implementation, The example value is 11, which ensures that enough data samples are covered to reflect local characteristics, while avoiding the loss of time series details caused by an excessively long window.

[0043] For the monitoring sequence corresponding to the c-th monitoring channel, each time t can be traversed along the time-series dimension. Taking time t as the center of the window, the sequence of that channel with length L centered at t is extracted. w A local data window, i.e., a subsequence containing data points at time t and their immediate vicinity, is used to accurately extract local data near time t. Furthermore, the arithmetic mean of all data points within the extracted local data window can be calculated to obtain the local average level of the c-th channel sequence at time t, denoted as [the average level]. , indicating the first The channel sequence with the first Centered on time, with a local window length of The arithmetic mean of the data within the window is used. This arithmetic mean can effectively suppress the interference of random fluctuations of a single data point, accurately depicting the local stationary benchmark and central tendency of the data in the c-th channel near time t. Furthermore, the above local data window can be slid along the time-series dimension according to a preset step size, and the above truncation and mean calculation operations can be repeated to generate the sequence corresponding to the c-th channel at each time step. This ultimately forms a local average level sequence with the same length as the original monitoring sequence.

[0044] Step S206: Based on the preset absolute safety boundary, determine the dynamic normalization boundary corresponding to the local average level.

[0045] Furthermore, for each monitoring channel and each moment, the upper and lower boundaries of dynamic normalization are calculated. These boundaries are determined by a predefined absolute safety hard boundary and the local statistical characteristics of the data at that moment. The preset absolute safety boundary is the safety threshold range of the physical quantities of each monitoring channel (such as upper / lower current limit, upper / lower voltage limit), which is the rigid bottom line to ensure the safe operation of the charging cabinet. Combining the aforementioned local average level, the normalization boundary that dynamically adjusts with the local average level can be calculated. For example, when the local average level is too high, the dynamic normalization boundary moves closer to the upper limit of the absolute safety boundary; when the local average level is too low, the dynamic normalization boundary moves closer to the lower limit of the absolute safety boundary. Based on this, the dynamic normalization boundary can both anchor the safety bottom line of the absolute safety boundary, ensuring that the normalization process respects physical constraints, and adapt to the local fluctuation characteristics of the data of each channel, achieving a balance between safety constraints and accurate data mapping.

[0046] In one implementation, it can be calculated using the following formula:

[0047]

[0048] In the formula, Indicates the first The absolute safety lower limit of the first channel is defined in advance according to the charging cabinet specifications. The absolute safety lower limit of each channel is a constant hard boundary that does not change with time. It is used to constrain the dynamic boundary to not be lower than this value. For example, the current channel can be set to 0 amps. Indicates the first The absolute safety limit for each channel is defined in advance according to the charging cabinet specifications. The absolute safety upper limit of each channel, as a constant hard boundary that does not change over time, is used to constrain dynamic boundaries from exceeding this value. For example, a voltage channel could be 300 volts. Indicates the channel index. Represents the current path. Represents the voltage channel. Represents the power channel; Indicates a point-in-time index. . Indicates the first The first channel in the The dynamic normalized lower boundary at time; Indicates the first The first channel in the The dynamic normalized upper bound of time. This represents the boundary scaling factor, a hyperparameter greater than 0, used to control the dynamic boundary bandwidth composed of local statistics. An example value is 2.0. This indicates the operation of retrieving the maximum value; This indicates the operation of taking the minimum value. Indicates the first The channel sequence with the first Centered on time, with a local window length of The standard deviation of the data within the window is used to characterize the local fluctuation amplitude of the data near that moment. The local window length is... An example value for is 11. Furthermore, the boundary can be extended using a symmetrically padded sequence to ensure that the window length is constant at each time step. This avoids statistical distortion.

[0049] Step S208: Based on the dynamic normalization boundary, the monitoring sequence of each monitoring channel is normalized to construct the target monitoring data corresponding to the running monitoring data.

[0050] The original data samples are normalized using the calculated dynamic normalization upper and lower boundaries, mapping their feature values ​​to the vicinity of the baseline interval. Notably, the normalized values ​​are allowed to exceed the range without strictly constraining the output value range. The range is defined, thus retaining out-of-bounds information as an anomaly signal, represented as:

[0051] In the formula, Represents normalized data No. The first channel in the The feature values ​​at each time point are transformed into relative values ​​based on the dynamic boundary through dynamic double-boundary normalization mapping. Represents the original data sample No. The first channel in the The value at each time point; This represents a very small positive number, used to prevent the denominator from being zero and to ensure numerical stability. Examples of possible values ​​are shown below. It should be noted that when When located within the dynamic boundary, The value is approximately interval, if Below the lower bound of dynamic normalization ,but ;like Above the upper bound of dynamic normalization ,but The mechanism for retaining out-of-bounds information can enhance the model's ability to identify abnormal behavior.

[0052] Furthermore, this embodiment of the invention also illustrates the working principle of the aforementioned boundary violation sensitive feature and its response to abnormal states, referring to... Figure 3This invention illustrates the working principle and response to abnormal states through a combination of three sub-graphs, visualizing the process of calculating the boundary violation sensitivity feature vector and demonstrating that this feature can effectively quantify abnormal risks and highlight violations near the boundary. The experimental analysis includes two scenarios: a normal charging current timing sequence and a charging current timing sequence containing an overcurrent anomaly. The lower safe boundary of the current is defined as 5 amps, and the upper safe boundary as 25 amps, marked with red dashed lines. The first sub-graph shows the fluctuation of the current within the safe boundary under normal conditions, with the green-filled area representing the safe operating range. The second sub-graph shows an overcurrent anomaly case where the current value exceeds the upper safe boundary during a period of 8 to 12 seconds; this "abnormal boundary crossing area" is clearly marked in red in the figure. The third sub-graph compares and calculates the boundary violation sensitivity feature value at each moment under the two states. Experimental results show that under the normal state in the first sub-graph, the boundary violation sensitivity feature value curve (blue) remains at a low level and fluctuates gently. In the abnormal state of the second subgraph, the corresponding eigenvalue curve (red) is close to the blue curve before and after the normal 8 seconds and 12 seconds. However, in the abnormal boundary crossing region (8 to 12 seconds), the red curve rises sharply, forming a significant peak. This indicates that when the current exceeds the absolute safety boundary, the boundary violation sensitive eigenvalue calculated in this embodiment of the invention will produce a strong positive response. More importantly, this feature is not only sensitive to explicit boundary crossings, but the boundary proximity perception term can also quantify the degree to which data points approach the boundary, providing early warning of potential risks.

[0053] Step S210: Construct the channel interaction fusion vector of the target monitoring data and determine the boundary violation sensitive features of the target monitoring data.

[0054] Although the normalized data of current, voltage, and power time series data has retained physical boundary information, the three channels remain independent. The operating state of the charging cabinet is essentially determined by the coupled dynamics of multiple physical quantities. Conventional feature fusion methods use simple channel splicing, which cannot uncover the deep spatiotemporal coupling relationship between channels, making it difficult for the model to capture the joint feature patterns that determine the operating state.

[0055] To address this, the present invention constructs a corresponding channel interaction fusion vector. Specifically, this embodiment employs a coupling tensor enhancement method, introducing a learnable coupling kernel tensor to model and enhance the interaction information between channels, generating a feature tensor (i.e., an intermediate tensor) rich in coupling relationships. Further, adaptive singular value filtering is used for denoising and focusing, extracting robust and significant joint feature patterns. Correspondingly, this embodiment constructs the channel interaction fusion vector through the following steps: 1) Use the preset coupling kernel tensor to perform inter-channel information fusion on the monitoring sequence of each monitoring channel of the target monitoring data to generate the intermediate tensor corresponding to the target monitoring data.

[0056] The normalized data mentioned above is a 3-channel single-depth feature. In this embodiment of the invention, it is further transformed into a 3-channel feature. Deep feature enhancement aims to specifically mine feature information at different levels, enriching the dimensions of feature representation. For example, it can enhance features with dimensions of... Normalized data Reconstructing into dimensions is A third-order tensor, with the last dimension initialized to 1; simultaneously, the augmented depth dimension is set to... , represents the number of coupling patterns to be mined, is a preset hyperparameter, and its value determines the number of potential coupling feature patterns mined and generated from the original three-channel data. An example value is 32.

[0057] Furthermore, by introducing a trainable coupled kernel tensor, the original 3-channel single-depth features can be transformed into 3-channel D-depth enhanced features through tensor multiplication operations. This allows information from each channel to interact and fuse with other channels at multiple depths, generating an intermediate tensor, represented as:

[0058] In the formula, The intermediate tensor, through learnable coupling kernels that fuse inter-channel information, represents the joint representation of multi-channel time-series data across multiple depth modes, with a dimension of [missing information]. . Represents intermediate tensor In the Time, Number The vector at depth is a 3-dimensional column vector, representing the coupled feature vectors of the three channels at a specific time and depth, i.e., the first... Joint representation of various coupling modes. The coupled kernel tensor is a trainable parameter with dimension O(n). The coupling relationship between channels can be modeled using tensor multiplication. Represents the coupled kernel tensor The middle corresponds to the input channel and deep mode A 3D column vector representing the original first... The enhanced three channels are all in the first... Contribution weights in deep mode. Indicates the first The bias vector corresponding to the depth is a trainable parameter. Among them, Corresponding to the third-order tensor of the above reconstruction .

[0059] 2) Perform singular value decomposition on the intermediate tensor to determine the first change pattern of the target monitoring data with respect to the joint time dimension and the second change pattern with respect to the data depth dimension.

[0060] 3) Based on the first singular vector matrix corresponding to the first change mode and the second singular vector matrix corresponding to the second change mode, perform data filtering on the target monitoring data to generate the channel interaction fusion vector of the target monitoring data.

[0061] Singular value decomposition (SVD) can extract key feature patterns and suppress noise interference from high-dimensional coupled intermediate tensors, enabling precise hierarchical and quantitative representation of feature information at different levels. The intermediate tensor integrates three dimensions of information from charging cabinet monitoring data: channel dimension (3 physical quantities), time dimension (T time series points), and depth dimension (D-layer enhanced features). Data from different dimensions are coupled, making direct analysis susceptible to noise interference and difficult to locate core patterns. To extract the most significant coupling patterns from the intermediate tensor and suppress noise, it is first reshaped into a two-dimensional matrix to map the high-dimensional coupling information to a unified two-dimensional space. Then, economical SVD is performed, decomposing the matrix into three parts: a left singular vector matrix, a singular value matrix, and a right singular vector matrix, thus decoupling the cross-dimensional coupling patterns. Let represent the left singular vector matrix, characterizing the main variation pattern of the data along the joint channel-time dimension (i.e., the first variation pattern, such as the coordinated fluctuation pattern of current and voltage over time). Its column vectors form an orthogonal basis with dimension . ; Let represent a right singular vector matrix, whose column vectors span a space that characterizes the primary variation pattern of the data in the depth dimension (i.e., the secondary variation pattern, such as the weight distribution pattern of augmented features at different depths). Its column vectors form an orthogonal basis with dimension . .

[0062] Furthermore, This represents a singular value matrix, which can be used to screen key patterns and suppress noise redundancy based on singular value quantization. This matrix is... The diagonal matrix, i.e. Its diagonal elements are singular values ​​and satisfy the following conditions: The magnitude of singular values ​​indicates the importance of the corresponding pattern: the first few larger singular values ​​correspond to the most significant core feature patterns in the data, while the last few smaller singular values ​​mostly correspond to noise or redundant information. Indicates the first The singular values ​​are positive and represent the importance or energy contribution of the pattern represented by the corresponding singular vector in the data.

[0063] Of the parameters mentioned above, Represents the following defined two-dimensional matrix The rank of the singular value decomposition (SVD) is set to r (much smaller than the original matrix dimension). This method retains only the first r key singular values ​​and their corresponding singular vectors, filtering out noise interference, reducing data dimensionality, and accurately preserving core feature patterns at both the channel-time and depth levels, thus avoiding irrelevant information from interfering with subsequent state recognition. Correspondingly, the two-dimensional matrix can be represented as: In the formula, The intermediate two-dimensional matrix is ​​represented by the intermediate tensor. The reshaped two-dimensional matrix has dimensions of The reshaping operation merges the channel and time dimensions into rows, and the depth dimension into columns; Describes a left singular vector matrix. This represents a right singular vector matrix. express The transpose of .

[0064] Furthermore, soft thresholding filtering can be applied to the singular values ​​obtained from the decomposition to attenuate small singular values ​​that may correspond to noise, while retaining large singular values ​​that correspond to the main coupling modes, as follows:

[0065] In the formula, Indicates the th after soft thresholding A number of singular values ​​are used to construct a new singular value matrix, thereby attenuating noise patterns and preserving the dominant patterns when reconstructing the filter matrix. It is a symbolic function, because Therefore ; This represents an adaptive soft threshold, used to control the shrinkage amount in soft threshold filtering. The calculation method is expressed as follows: . represents the scaling factor, a hyperparameter controlling the filtering intensity, with an example value of 0.6. In summary, this process attenuates smaller singular values ​​by preserving the sign of the singular values ​​and applying a soft threshold function using a maximum value comparison method.

[0066] Furthermore, a new diagonal matrix can be constructed using the singular values ​​obtained after soft thresholding. This matrix is ​​then used in conjunction with the original left and right singular vector matrices to obtain the reconstructed filter matrix after denoising and focusing, as follows: In the formula, This represents the reconstruction of the filter matrix, which removes singular values ​​corresponding to noise and highlights the main coupling modes, thus obtaining the denoised and focused feature representation with dimension . . This represents a diagonal matrix whose diagonal elements are the singular values ​​after soft thresholding. .

[0067] Furthermore, the above-mentioned reconstructed filter matrix can be... Reconstructing it into a three-dimensional format yields the enhanced feature tensor. (That is, the channel interaction fusion vector), with dimensions of This enhanced feature tensor, based on the original normalized features, adds two new dimensions of information: cross-channel temporal correlation features and deep hierarchical weighted features. This upgrades the feature representation from single-dimensional temporal data to multi-dimensional, hierarchical, and weighted structured features, significantly enriching the feature representation dimensions and improving the feature discrimination between normal and abnormal states (especially compound anomalies and early critical anomalies). In summary, the embodiments of this invention can construct a channel interaction fusion vector that not only retains the original temporal information of each channel but also embeds a significant inter-channel coupling pattern after adaptive filtering.

[0068] The operating status of a charging cabinet depends not only on instantaneous features but also on its temporal evolution and the topological pattern formed by multiple features. Conventional classification methods typically flatten features and directly input them into a fully connected network for mapping, ignoring the temporal dependence of features and their internal geometric structure. Furthermore, they lack targeted modeling for the differences in boundary sensitivity required for different state discriminations, resulting in insufficient generalization and identification capabilities for complex anomaly patterns. To address this, this invention proposes a state classification method that integrates violation-sensitive gating, spatiotemporal attention focusing, and multi-prototype metric learning to achieve accurate and robust identification of the charging cabinet's operating status. For the corresponding violation-sensitive gating, this embodiment calculates the boundary violation-sensitive feature vector corresponding to the data, quantifying the anomaly risk of different channels at different times, thereby enabling the model to be alert to physical safety boundaries.

[0069] In specific implementation, based on preset boundary crossing parameters (in one implementation, the aforementioned absolute safety boundary can be used as the boundary crossing parameter), the boundary crossing event data corresponding to the target monitoring data can be determined; and based on the normalized relative boundary corresponding to the normalization processing, the boundary proximity data corresponding to the target monitoring data can be determined. Then, based on the boundary crossing event data and the boundary proximity data, the boundary violation sensitive features of the target monitoring data are determined. Combining the above steps, this embodiment of the invention utilizes the aforementioned absolute safety boundary and local statistics to construct corresponding boundary violation sensitive features to capture explicit absolute boundary crossing behavior and quantify the proximity of data points to the safety boundary. This can be expressed as:

[0070] In the formula, Indicates the first The first channel, the first The boundary violation sensitive characteristic value at a given time is a scalar, and its value directly reflects the level of anomaly risk at that location. The larger the value, the higher the risk. Indicates the first The channel sequence with the first Standard deviation within a local window centered on time; This represents the numerical stabilization increment, a very small positive number used to prevent the denominator from being zero. Examples of possible values ​​are: . This represents the natural exponential function. This represents the proximity perception weight coefficient, a hyperparameter used to balance the contributions of the absolute out-of-bounds penalty term and the boundary proximity perception term. An example value is 0.5. This represents the hyperbolic tangent activation function, which maps the input to... The interval is used to smooth and saturate amplify signals that are close to the boundary. Indicates the first The first channel in the The normalized relative boundary proximity at time step is used to quantify the relative distance of a data point from the nearest absolute boundary. The smaller the value, the closer it is to the boundary. The calculation method is expressed as follows:

[0071] Indicates the first The first channel in the The positive distance from the absolute safety upper boundary at any given time is used to measure the degree of boundary breach, and is calculated as follows: ,when The time indicates that an upper bound has occurred, and can be used to indicate the occurrence of a corresponding bounds overflow event; Indicates the first The first channel in the The positive distance from the absolute safety lower boundary at any given time is used to measure the degree of boundary breach, and is calculated as follows: ,when The time indicates that a lower bound has occurred, and can be used to indicate the occurrence of a corresponding boundary overflow event. Represents the original data sample In the The first channel, the first The value at each time point.

[0072] Step S212: The boundary violation sensitive features are fused into the channel interaction fusion vector to construct the focused feature tensor corresponding to the target monitoring data.

[0073] The channel interaction fusion vector is a global correlation feature extracted based on the channel-time-depth cross-dimensional coupling mode, representing the collaborative change law of multiple physical quantities; the boundary violation sensitive feature is a local critical feature, representing the abnormal precursor of data approaching the safety threshold. This embodiment of the invention fuses the two, so that the corresponding features simultaneously contain feature information at both the global correlation level and the local critical level. This allows for focusing on the differential information of the two types of features, amplifying the feature differences between normal and abnormal states (especially early critical anomalies and complex collaborative anomalies), thereby achieving highly discriminative classification and prediction.

[0074] In this embodiment of the invention, a preset gating fusion mechanism is used to fuse boundary violation sensitive features into the channel interaction fusion vector to construct an initial fusion feature tensor. Based on preset multi-dimensional attention weights, the initial fusion feature tensor is adaptively weighted to construct a focused feature tensor. In one implementation, a channel- and time-adaptive gating fusion mechanism can be used to inject boundary risk perception information into the enhanced feature tensor to obtain the fusion feature tensor. This allows the model to dynamically adjust the contribution of the enhanced features based on the risk value at each location and directly inject risk information. Specifically, the calculation method of the fusion feature tensor is expressed as follows:

[0075] In the formula, Represents the fusion feature tensor In the The first channel, the first Time, Number The feature values ​​of each deep pattern represent the deep coupling features (i.e., the initial fused feature tensor) after being modulated by boundary risk perception and injected with direct risk signals. The dimension is . This represents the above-mentioned enhanced feature tensor In the The first channel, the first Time, Number Feature values ​​of a depth pattern; This represents the element-wise multiplication operator; This represents the Sigmoid activation function, which maps the input to... The interval is used as the gating weight. Indicates the first The first channel, the first The gating transformation weights corresponding to each depth pattern are trainable parameters, which are scalars used to map boundary violation sensitive feature values ​​to gating signals. Indicates the first The gating bias corresponding to each depth pattern is a trainable parameter and is a scalar. Indicates the first The first channel, the first The risk information injection weights corresponding to each deep pattern are trainable parameters, which are scalars used to linearly transform the boundary violation sensitive features and then inject them into the enhanced feature tensor.

[0076] Furthermore, the contribution of different stages of charging cabinet operation and different coupling modes to the final state determination varies. To focus on the most critical time segments and the most significant coupling modes, this embodiment of the invention also employs a dual attention mechanism to simultaneously calculate the attention weights in the time and depth dimensions without flattening the tensor, expressed as follows:

[0077] In the formula, Represents the attention weight matrix In the The first channel, the first Time, Number Each element of a deep pattern has a numerical value that represents the importance score of the corresponding feature. Indicates the first The first channel, the first Time, Number The attention score for each deep pattern is calculated as follows:

[0078] Indicates the first The first channel, the first The query vector at time step is obtained by linearly transforming all depth feature vectors of the fused feature tensor at a specific channel and time point. The calculation method is expressed as follows: . Represents the fusion feature tensor In the The first channel, the first The feature vectors at all times and in all deep modes have a dimension of ; The weight matrix represents the query transformation and is a trainable parameter used to linearly project the feature vectors into a new vector space. Indicates the first The first channel, the first The key vector of the depth pattern, through the first depth pattern... The first channel, the first All time points in the depth model The average of the eigenvalues ​​is used to represent the global temporal summary of this channel-depth combination, and the calculation method is expressed as follows:

[0079] The weight matrix representing the key transformation is a trainable parameter used to linearly project the pooled feature vectors onto a new vector space. This represents the dimension of the key vector, set by default. ; The content augmentation coefficient is a trainable scalar parameter used to adjust the degree of direct influence of the original feature values ​​on the attention weights.

[0080] Furthermore, the initial fused feature tensor is reweighted using the aforementioned attention weight matrix to highlight feature elements that contribute more to state recognition and suppress unimportant features, as shown below:

[0081] In the formula, Represents the focused feature tensor In the The first channel, the first Time, Number The feature values ​​of a deep pattern represent the deep coupling features (i.e., the focused feature tensor) that have been calibrated for importance and carry boundary risk awareness.

[0082] Step S214: Use the pre-built operation status monitoring model to classify and predict the focused feature tensor to determine the operation status classification result corresponding to the target charging cabinet.

[0083] After determining the corresponding focused feature tensor, this embodiment of the invention uses a pre-built model to perform classification prediction based on the learned feature patterns, and outputs the corresponding operating state classification results. For example, these include normal operating state, critical warning state, and specific anomaly type (overvoltage / overcurrent, etc.).

[0084] Considering that different operating states may correspond to multiple typical patterns in the feature space, this embodiment of the invention employs multi-prototype metric learning for classification output. In specific implementation, a multi-prototype metric learning mechanism is used to determine the probability distribution prediction result of each monitoring channel of the focused feature tensor based on its corresponding pattern summary vector; based on the probability distribution prediction result, the operating state classification result corresponding to the target charging cabinet is determined. Specifically, the focused feature tensor is first aggregated into a pattern summary vector for each channel, then the minimum Mahalanobis distance between it and multiple prototypes of each state category is calculated, and finally, it is converted into a probability distribution to obtain the predicted probability that the sample belongs to each operating state category, expressed as:

[0085] In the formula, Indicates that the sample belongs to the first The predicted probability of each running state category, with the sum of the probabilities of all categories being 1. Indicates the index of running status category. ; This indicates the total number of categories of operating status. For example, specific categories include: 1-Normal, 2-Overvoltage, 3-Undervoltage, 4-Overcurrent, 5-Power anomaly, and 6-Combined anomaly. Indicates the first The number of prototypes corresponding to each category is a hyperparameter; Indicates the prototype index. Indicates the selection of the first The operation is to find the minimum nearest distance among all prototypes in each category.

[0086] Indicates sample features and the first The first category The overall distance between prototypes is calculated as follows: ; The calculation method for the squared Mahalanobis distance is as follows: ;in It is a trainable, positive semi-definite diagonal weight matrix used to measure the difference in importance of different depth pattern dimensions. yes The transpose, and It is a column vector. Indicates the first Category 1, No. 2 The corresponding prototype in the first The prototype subvectors of each channel are trainable parameters with dimension . . Indicates the first The pattern summary vector for each channel can be obtained by averaging the focus feature tensor over the time dimension, and the calculation method is expressed as follows: , dimension . Represents the focused feature tensor In the The first channel, the first Feature vectors at all times and across all depth modes.

[0087] In summary, for the target charging cabinet in this embodiment of the invention, the real-time or offline operating status monitoring of one or more charging cabinets is first performed, and the data acquisition process is initiated (i.e., through the IoT sensor network of the target charging cabinet, current or recent time-series monitoring data of current, voltage, and power are synchronously collected according to the same specifications as in the training phase, forming a raw data sample to be detected). Then, the sample is processed using the aforementioned dynamic dual-boundary normalization method, utilizing predefined absolute safety boundaries and the same local statistical calculation rules to convert it into normalized data, thereby preprocessing the data and highlighting possible boundary violation information. Furthermore, channel interaction fusion and boundary violation sensitive features can be fused into the aforementioned model to classify and identify the normalized data through the model. The model has been trained and loaded with optimal parameters. Internally, the model sequentially performs multi-source time-series feature coupling tensor enhancement (i.e., constructing a channel interaction fusion vector), and classifies the charging cabinet operating status based on the determined boundary violation sensitive features.

[0088] Corresponding to the above embodiments, the model first mines deep coupling patterns between current, voltage, and power sequences and filters out noise through a coupling tensor enhancement module. Then, it combines the boundary violation sensitive features corresponding to the original data through a classification module, focuses key information through gating fusion and spatiotemporal attention mechanisms, and finally calculates the probability that the sample to be detected belongs to each preset operating state category (normal, overvoltage, undervoltage, overcurrent, power anomaly, and combined anomaly) through multi-prototype metric learning. The model output is a probability distribution vector, where the category with the highest probability is determined as the operating state of the charging cabinet during the current monitoring period. Based on this determination result, the monitoring system can trigger corresponding early warnings, alarms, or log records in real time, thereby realizing automated and intelligent monitoring of the charging cabinet's operating status and early fault identification, providing direct decision support for operation and maintenance management. Correspondingly, Figure 4 A schematic diagram of the classification prediction process corresponding to an embodiment of the present invention is shown.

[0089] Furthermore, in conjunction with the above embodiments, the operational status monitoring model of this invention can be constructed based on the channel interaction fusion vector and boundary violation sensitive features of the corresponding training samples, expanding the model's feature representation dimension from a single physical quantity change dimension to a two-dimensional space of global coupling + local criticality. On the one hand, the global coupling feature ensures the model's ability to identify multi-parameter coordinated anomalies, avoiding the missed detection problem of single-parameter monitoring; on the other hand, the local criticality feature enhances the model's sensitivity to early fault precursors, breaking the limitation of post-event judgment that alarms only when the standard is exceeded. The combination of the two can give the model's feature space higher discriminative power, accurately distinguishing between normal operation, critical warning, and specific anomaly types (overvoltage, overcurrent, etc.), significantly improving the identification accuracy and generalization ability of charging cabinet operational status under complex working conditions. Correspondingly, after the model is trained, its output results can include both the classification of operational status and the ability to trace back and locate the root cause of anomalies (such as a critical anomaly of a certain channel or a coordinated anomaly of multiple channels) through feature weights. This enables the model not only to achieve automatic status identification but also to provide maintenance personnel with accurate anomaly location basis.

[0090] Based on the aforementioned monitoring data, corresponding training samples can be constructed to form a standardized dataset for training the charging cabinet operation status monitoring model. This involves collecting raw time-series monitoring sequences from a large number of historical operation records of charging cabinets in normal and various abnormal states to construct the corresponding training dataset. The data acquisition process must ensure sensor accuracy, sampling rate consistency, and strict timestamp synchronization. After data acquisition, each raw data sample is precisely labeled to provide the supervision signals required for model training. The labeling work can be based on historical fault records, maintenance logs, and expert knowledge to determine and classify the charging cabinet operation status corresponding to each time-series data sample. In one implementation, six operation status categories are defined, specifically including: normal state, overvoltage state, undervoltage state, overcurrent state, power abnormal state, and composite abnormal state. A composite abnormal state refers to a situation that simultaneously contains multiple single abnormal features or presents a complex abnormal pattern. The labeling results are given in the form of sample labels using a one-hot encoding format. Finally, a large number of labeled raw data samples and their corresponding one-hot encoded labels can be combined to form a complete training dataset.

[0091] Furthermore, the loss function employed in this embodiment of the invention not only supervises the correctness of the final classification but also guides the model to focus more on high-risk samples and ensures the stability of feature enhancement by introducing violation-aware sample weights and constraints on the coupled-mode filtering process. In specific implementation, the loss function includes a violation-aware weighted classification loss and a coupled-mode filtering consistency regularization loss. The violation-aware weighted classification loss is calculated based on the dynamic weights of the training samples, and the dynamic weights are determined based on boundary violation-sensitive features; the coupled-mode filtering consistency regularization loss is used to constrain the consistency of the filtering parameters corresponding to the channel interaction fusion vector.

[0092] 1) Classification loss weighted by violation perception: Cross-entropy loss is fundamental to supervised learning, but abnormal samples in charging cabinets are usually few in number and highly dangerous. To improve the model's classification accuracy for high-risk samples, a dynamic weight is assigned to each training sample based on the boundary violation sensitivity feature, thus constructing a violation-aware weighted classification loss, expressed as:

[0093] In the formula, This represents the violation-aware weighted classification loss, which serves as the primary supervision signal during model training to ensure correct model classification. This represents the total number of samples in the training batch; Indicates the sample index; Indicates the first The dynamic weights of each sample are calculated as follows: This is used to amplify the loss contribution of high-risk samples. Indicates traversing the first... For each sample, take the maximum value of the boundary violation sensitive feature value among all channels and all time points, and use it to extract the highest violation risk value of the sample among all channels and all time points; This represents the violation weighting enhancement coefficient, a hyperparameter greater than 0, used to control the sensitivity of sample weights to the degree of violation. An example value is 1.0. Indicates the first The sample at the th The first channel, the first Boundary violation sensitive feature value at any given time; Indicates the first The true label of the sample is obtained using one-hot encoding. When the sample belongs to the... The value is 1 if the condition is met, otherwise it is 0. Indicates the first The sample belongs to the first The predicted probability of each running state category; This represents a logarithmic function, with the default base being the natural constant.

[0094] 2) Regarding the consistency regularization loss of coupled-mode filtering: Adaptive singular value filtering aims to remove noise and focus on key coupling modes. To avoid the filtering process being too drastic and losing useful information or introducing instability, this embodiment of the invention also employs coupling mode filtering consistency regularization loss to constrain the filtered singular values ​​from deviating excessively from the original singular values. Simultaneously, it encourages the sparsity of the coupling kernel tensor to improve interpretability, expressed as follows:

[0095] In the formula, This represents the coupling mode filter consistency regularization loss, used to constrain and guide the feature enhancement process, encouraging the learned inter-channel coupling relationships to be more concise and interpretable. This represents the singular value consistency regularization coefficient, a hyperparameter greater than 0, used to control the penalty strength for the difference in singular values ​​before and after filtering. An example value is 0.1. This represents the sparsity regularization coefficient of the coupling kernel. It is a hyperparameter greater than 0 and is used to control the penalty strength on the absolute value of the coupling kernel tensor elements. An example value is 0.01. Represents a trainable coupled kernel tensor In the original channel Output channels and deep mode The scalar weights at positions represent the positional weights in the generation of the first position. When considering the features of a deep pattern, from the original... Information from the first channel is transmitted to the enhanced first channel. The weights of each channel are determined. By weighted summing of the contributions of all original channels, information exchange and fusion between channels are achieved. Indicates difference from Channel index.

[0096] 3) Calculate the total loss function: The model's total loss function is the sum of the violation-aware weighted classification loss and the coupled-mode filtering consistency regularization loss, i.e., the total loss function. Alternatively, a weighted summation method can be used to calculate the total loss function. ,For example, The weighted values ​​are preset hyperparameters.

[0097] Furthermore, the corresponding model training is based on the training dataset constructed in the aforementioned steps. The model prediction results are calculated using forward propagation, and the aforementioned total loss function is used as the performance metric and optimization objective. The training process can employ a gradient descent-based backpropagation algorithm, iteratively optimizing and updating all trainable parameters in the model. At the start of training, all trainable parameters in the model are initialized. In each training iteration (epoch) or mini-batch, a batch of samples is extracted from the training dataset, sequentially processed through the aforementioned dynamic double-boundary normalization preprocessing, and then input into the monitoring model for forward propagation. The forward propagation process follows the aforementioned data processing and classification output steps, sequentially executing multi-source temporal feature coupling tensor enhancement and violation-sensitive gating fusion, spatiotemporal attention focusing, and multi-prototype metric classification, ultimately outputting the predicted probability of each sample belonging to each running state category. Further, based on the predicted probability and the sample's true label, the aforementioned total loss function value can be calculated. This value combines the violation-aware weighted classification loss and the coupled-pattern filtering consistency regularization loss. The gradient of the loss function with respect to all trainable parameters is calculated, and these parameters are updated using a selected optimizer (such as the Adam optimizer) based on hyperparameters such as gradient direction and learning rate, thereby minimizing the total loss function. This process is repeated until a preset stopping condition is met. Further, the stopping condition can be set to the classification performance (such as accuracy) on the independent validation set no longer improving and persisting for a certain number of iterations (early stopping), or reaching a preset maximum number of training iterations. When training ends, the optimal model parameters are saved, thus completing the training and parameter update of the charging cabinet operation status monitoring model.

[0098] In summary, the embodiments of the present invention are innovative compared to the prior art in the following aspects: 1. A dynamic dual-boundary normalization method is adopted. By combining the absolute safety boundary and local statistics, a time-varying normalized boundary is constructed, and out-of-boundary values ​​are retained outside the baseline interval, making the model sensitive to boundary violations and near-boundary behavior.

[0099] 2. A multi-source temporal feature coupling tensor enhancement module is adopted, which introduces a learnable coupling kernel tensor to model the interaction relationship between multiple physical quantities, and combines adaptive singular value filtering to extract robust coupling modes, thus solving the problem that simple feature splicing is difficult to mine deep spatiotemporal correlations.

[0100] 3. A classification mechanism that integrates violation perception gating, spatiotemporal dual attention, and multi-prototype metric learning is adopted. The abnormal risk is quantified by boundary violation sensitive features, and adaptive feature calibration is achieved by gating fusion and attention focusing. Finally, the ability to identify complex abnormal patterns is enhanced by multi-prototype matching.

[0101] 4. By adopting the violation-aware weighted loss function and the coupled-mode filtering consistency regularization loss, the model strengthens its focus on high-risk abnormal samples through dynamic sample weights, and constrains the stability and interpretability of the feature enhancement process, thereby improving the model's performance in imbalanced data scenarios.

[0102] Furthermore, based on the above embodiments, this invention also provides a charging cabinet operation status monitoring device based on Internet of Things (IoT) data, referring to... Figure 5 The device includes: a data monitoring module 10, used to determine the operation monitoring data of the target charging cabinet through the IoT sensor network of the target charging cabinet; a data preprocessing module 20, used to normalize the operation monitoring data based on a preset absolute safety boundary and the local statistical features corresponding to the monitoring sequence of each monitoring channel in the operation monitoring data, to generate target monitoring data; a data processing module 30, used to construct the channel interaction fusion vector of the target monitoring data and determine the boundary violation sensitive features of the target monitoring data; and an execution module 40, used to classify and predict the target monitoring data based on the channel interaction fusion vector and the boundary violation sensitive features, to determine the operation status classification result of the target charging cabinet.

[0103] The charging cabinet operation status monitoring device based on Internet of Things data provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0104] Furthermore, this embodiment of the invention also provides a charging cabinet operation status monitoring system based on Internet of Things data. This system is configured with the apparatus described in the above embodiments and is used to execute the methods of any of the above embodiments.

[0105] The charging cabinet operation status monitoring system based on Internet of Things data provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0106] This invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described... Figures 1 to 2 The steps of any of the methods shown. Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the above-described steps. Figures 1 to 2 The steps of any of the methods shown. Embodiments of the present invention also provide a structural schematic diagram of an electronic device, such as... Figure 6The diagram shows the structure of the electronic device, which includes a processor 101 and a memory 100. The memory 100 stores computer-executable instructions that can be executed by the processor 101. The processor 101 executes the computer-executable instructions to implement the above-mentioned... Figures 1 to 2 Any of the methods shown.

[0107] exist Figure 6 In the illustrated embodiment, the electronic device further includes a bus 102 and a communication interface 103, wherein the processor 101, the communication interface 103, and the memory 100 are connected via the bus 102. The memory 100 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk drive. Communication between this system network element and at least one other network element is achieved through at least one communication interface 103 (which can be wired or wireless), using the Internet, wide area network, local area network, metropolitan area network, etc. Bus 102 can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, or an AMBA (Advanced Microcontroller Bus Architecture) bus. AMBA defines three types of buses: APB (Advanced Peripheral Bus), AHB (Advanced High-performance Bus), and AXI (Advanced eXtensible Interface). Bus 102 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6The diagram uses only a single double-headed arrow, but this does not imply a single bus or a single type of bus. Processor 101 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 101 or by instructions in software form. Processor 101 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor 101 reads information from the memory and, in conjunction with its hardware, completes the aforementioned tasks. Figures 1 to 2 Any of the methods shown.

[0108] The computer program product of the charging cabinet operation status monitoring method, device, and system based on Internet of Things data provided in this invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the preceding method embodiments, which will not be repeated here. In addition, in the description of the embodiments of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "connection" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, 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 described in the various embodiments of the present invention. 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.

[0109] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., 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 the invention and for simplifying the description, and do not 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 the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Finally, it should be noted that the above embodiments are merely specific implementations of the invention, used to illustrate the technical solutions of the invention, and not to limit it. The scope of protection of the invention is not limited thereto. Although the invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in this invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be covered within the scope of protection of this invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for monitoring the operating state of an Internet of Things data-based charging cabinet, characterized in that, The method comprises: determining operation monitoring data of a target charging cabinet through an Internet of Things sensor network of the target charging cabinet; based on a preset absolute safety boundary and local statistical features corresponding to each monitoring channel in the operation monitoring data, performing normalization processing on the operation monitoring data to generate target monitoring data; constructing a channel interaction fusion vector of the target monitoring data and determining a boundary violation sensitive feature of the target monitoring data; based on the channel interaction fusion vector and the boundary violation sensitive feature, performing classification prediction on the target monitoring data to determine an operation state classification result of the target charging cabinet.

2. The method of claim 1, wherein, The step of performing normalization processing on the operation monitoring data based on a preset absolute safety boundary and local statistical features corresponding to each monitoring channel in the operation monitoring data to generate target monitoring data comprises: based on a preset local window length, determining a local average level corresponding to each monitoring channel in the operation monitoring data; based on a preset absolute safety boundary, determining a dynamic normalization boundary corresponding to the local average level; based on the dynamic normalization boundary, performing normalization processing on the monitoring sequence of each monitoring channel to construct target monitoring data corresponding to the operation monitoring data.

3. The method of claim 1, wherein, The step of constructing a channel interaction fusion vector of the target monitoring data comprises: using a preset coupling kernel tensor to perform inter-channel information fusion on the monitoring sequence of each monitoring channel of the target monitoring data to generate an intermediate tensor corresponding to the target monitoring data; performing singular value decomposition on the intermediate tensor to determine a first change mode for a time joint dimension and a second change mode for a data depth dimension of the target monitoring data; based on a first singular vector matrix corresponding to the first change mode and a second singular vector matrix corresponding to the second change mode, performing data filtering processing on the target monitoring data to generate a channel interaction fusion vector of the target monitoring data.

4. The method of claim 1, wherein, The step of determining a boundary violation sensitive feature of the target monitoring data comprises: based on a preset out-of-bound parameter, determining out-of-bound event data corresponding to the target monitoring data; based on a normalization relative boundary corresponding to the normalization processing, determining boundary proximity data corresponding to the target monitoring data; based on the out-of-bound event data and the boundary proximity data, determining a boundary violation sensitive feature of the target monitoring data.

5. The method of claim 1, wherein, The step of performing classification prediction on the target monitoring data based on the channel interaction fusion vector and the boundary violation sensitive feature to determine an operation state classification result of the target charging cabinet comprises: fusing the boundary violation sensitive feature into the channel interaction fusion vector to construct a focused feature tensor corresponding to the target monitoring data; using a pre-constructed operation state monitoring model to perform classification prediction on the focused feature tensor to determine an operation state classification result corresponding to the target charging cabinet.

6. The method of claim 5, wherein, The step of fusing the boundary violation sensitive feature into the channel interaction fusion vector to construct the focused feature tensor corresponding to the target monitoring data comprises: The preset gating fusion mechanism is adopted to fuse the boundary violation sensitive feature into the channel interaction fusion vector to construct an initial fusion feature tensor; The initial fusion feature tensor is adaptively weighted based on a preset multi-dimensional attention weight to construct a focused feature tensor.

7. The method of claim 5, wherein, The step of using a pre-constructed operation state monitoring model to perform classification prediction on the focused feature tensor to determine the operation state classification result corresponding to the target charging cabinet comprises: A multi-prototype metric learning mechanism is adopted to determine that each monitoring channel of the focused feature tensor is based on a probability distribution prediction result of a corresponding mode summary vector; Based on the probability distribution prediction result, the operation state classification result corresponding to the target charging cabinet is determined.

8. The method of claim 5, wherein, The operation state monitoring model is constructed based on the channel interaction fusion vector and the boundary violation sensitive feature of the preset training sample; The loss function of the operation state monitoring model comprises a violation perception weighted classification loss and a coupled mode filtering consistency regularization loss; The violation perception weighted classification loss is calculated based on a dynamic weight of the training sample, and the dynamic weight is determined based on the boundary violation sensitive feature; The coupled mode filtering consistency regularization loss is used to constrain the consistency of the filtering parameters corresponding to the channel interaction fusion vector.

9. A charging cabinet operation state monitoring device based on Internet of Things data, characterized in that, The device comprises: A data monitoring module configured to determine operation monitoring data of a target charging cabinet through an Internet of Things sensor network of the target charging cabinet; A data preprocessing module configured to perform normalization processing on the operation monitoring data based on a preset absolute safety boundary and local statistical features corresponding to a monitoring sequence of each monitoring channel of the operation monitoring data to generate target monitoring data; A data processing module configured to construct a channel interaction fusion vector of the target monitoring data and determine a boundary violation sensitive feature of the target monitoring data; An execution module configured to perform classification prediction on the target monitoring data based on the channel interaction fusion vector and the boundary violation sensitive feature to determine an operation state classification result of the target charging cabinet.

10. A charging cabinet operation state monitoring system based on Internet of Things data, characterized in that, The charging cabinet operation state monitoring system based on Internet of Things data is configured with the charging cabinet operation state monitoring device based on Internet of Things data of claim 9.

Images