A remote monitoring method and system for a power utilization information collection terminal device and a medium

Abstract

The application discloses a kind of remote monitoring method, system and medium of electric information collection terminal equipment, it is related to electric parameter monitoring technical field, the method includes: remote monitoring terminal, obtains real-time acquisition data and monitors multidimensional sequence;Self-adapting analysis current node, constructs backtracking time zone and backtracks historical data;Based on backtracking data depth detection anomaly, obtain anomaly coefficient;If coefficient is over threshold, carry out multi-degree-of-freedom causal backtracking, generate anomaly backtracking atlas;If coefficient is not over threshold, execute multidimensional association implicit fault deduction, generate implicit fault atlas, to solve the technical problem that existing electric information collection terminal remote monitoring process is difficult to locate in time, implicit fault is difficult to effectively identify, reach by self-adapting backtracking feature analysis and multi-degree-of-freedom causal backtracking, improve terminal anomaly detection accuracy and fault backtracking reliability, realize the technical effect of efficient intelligent remote monitoring.

Description

Technical Field

[0001] This invention relates to the field of electrical parameter monitoring technology, and in particular to a remote monitoring method, system and medium for an electricity information acquisition terminal device. Background Technology

[0002] With the advancement of digitalization and IoT transformation of power distribution networks, a large number of electricity information collection terminal devices are distributed in substations, distribution areas, and user-side sites, carrying out functions such as power metering, event alarms, communication forwarding, and edge computing. The diverse sources of these terminals and the complexity of communication links, coupled with the significant impact of electromagnetic interference, temperature and humidity, power quality, and on-site construction on the operating environment, result in collected data exhibiting characteristics such as high dimensionality, multi-scale, strong noise, and intermittent gaps. Simultaneously, the obvious daily / weekly / seasonal nature of loads makes historical thresholds and static models prone to failure, and conceptual drift and clock drift increase the difficulty of anomaly detection. Existing remote operation and maintenance methods largely rely on fixed alarm rules or single time-series predictions, lacking sufficient sensitivity to identify sudden anomalies and gradual changes in potential hazards, and lacking interpretable traceability of anomaly causes across hardware, software, communication, and environmental dimensions, making early intervention difficult. Faced with large-scale terminal deployment and data deluge, traditional methods also suffer from efficiency and accuracy bottlenecks in retrospective evidence collection, correlation inference, and identification of small-sample hidden faults under imbalanced sample conditions, increasing the risk of unplanned outages. Summary of the Invention

[0003] This invention provides a remote monitoring method, system, and medium for electricity information collection terminal equipment to solve the technical problems of difficulty in timely location of anomalies and effective identification of hidden faults in the remote monitoring process of existing electricity information collection terminals. It achieves the technical effect of improving the accuracy of terminal anomaly detection and the reliability of fault tracing through adaptive backtracking feature analysis and multi-degree-of-freedom causal tracing, thereby realizing efficient and intelligent remote monitoring.

[0004] In a first aspect, the present invention provides a remote monitoring method for an electricity consumption information collection terminal device, wherein the remote monitoring method for the electricity consumption information collection terminal device includes:

[0005] A remote monitoring terminal device for electricity consumption information collection obtains terminal collection data and a multi-dimensional sequence of terminal monitoring corresponding to the current time node. Based on the multi-dimensional sequence of terminal monitoring, adaptive backtracking feature analysis is performed on the current time node to construct a backtracking feature time zone. The collection data from the electricity consumption information collection terminal device is then backtracked based on the backtracking feature time zone to obtain a backtracking collection dataset. Based on the backtracking collection dataset, the collection data undergoes a depth detection of collection anomalies to obtain a collection anomaly depth coefficient. If the collection anomaly depth coefficient is greater than or equal to a collection anomaly depth threshold, multi-degree-of-freedom causal tracing is performed on the collection anomaly depth coefficient based on the multi-dimensional sequence of terminal monitoring to obtain a terminal anomaly causal tracing map. If the collection anomaly depth coefficient is less than the collection anomaly depth threshold, multi-dimensional correlation latent fault inference is performed on the multi-dimensional sequence of terminal monitoring based on the backtracking feature time zone to obtain a terminal latent fault inference map.

[0006] Secondly, the present invention also provides a remote monitoring system for an electricity consumption information collection terminal device, wherein the remote monitoring system for the electricity consumption information collection terminal device includes:

[0007] Data Acquisition Module: Remotely monitors electricity consumption information acquisition terminal equipment to obtain terminal acquisition data and terminal monitoring multi-dimensional sequence corresponding to the current time node; Data Backtracking Module: Performs adaptive backtracking feature analysis on the current time node based on the terminal monitoring multi-dimensional sequence, constructs a backtracking feature time zone, and performs backtracking of the acquired data of the electricity consumption information acquisition terminal equipment based on the backtracking feature time zone to obtain a backtracking acquisition dataset; Acquisition Anomaly Detection Module: Performs acquisition anomaly depth detection on the terminal acquisition data based on the backtracking acquisition dataset to obtain an acquisition anomaly depth coefficient; Causal Tracing Module: If the acquisition anomaly depth coefficient is greater than or equal to the acquisition anomaly depth threshold, performs multi-degree-of-freedom causal tracing on the acquisition anomaly depth coefficient based on the terminal monitoring multi-dimensional sequence to obtain a terminal anomaly causal tracing map; Fault Inference Module: If the acquisition anomaly depth coefficient is less than the acquisition anomaly depth threshold, performs multi-dimensional correlation latent fault inference on the terminal monitoring multi-dimensional sequence based on the backtracking feature time zone to obtain a terminal latent fault inference map.

[0008] Thirdly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a remote monitoring method for an electricity information collection terminal device provided by the present invention.

[0009] This invention discloses a remote monitoring method, system, and medium for electricity information collection terminal equipment, comprising: remotely monitoring the electricity information collection terminal equipment to obtain terminal collection data and a terminal monitoring multidimensional sequence corresponding to the current time node; performing adaptive backtracking feature analysis on the current time node based on the terminal monitoring multidimensional sequence to construct a backtracking feature time zone, and performing backtracking of the collection data of the electricity information collection terminal equipment based on the backtracking feature time zone to obtain a backtracking collection dataset; performing collection anomaly depth detection on the terminal collection data based on the backtracking collection dataset to obtain a collection anomaly depth coefficient; if the collection anomaly depth coefficient is greater than or equal to a collection anomaly depth threshold, performing the terminal monitoring multidimensional sequence on the current time node to obtain a backtracking collection dataset; performing collection anomaly depth detection on the terminal collection data to obtain a collection anomaly depth coefficient; if the collection anomaly depth coefficient is greater than or equal to a collection anomaly depth threshold, performing the collection anomaly depth analysis on the current time node based on the terminal monitoring multidimensional sequence to obtain a backtracking collection dataset; and performing collection anomaly depth detection on the terminal collection data based on the backtracking collection dataset to obtain a collection anomaly depth coefficient. The method, system, and medium for remote monitoring of electricity information collection terminal equipment disclosed in this invention solve the technical problems of difficulty in timely location of anomalies and effective identification of hidden faults in the remote monitoring process of existing electricity information collection terminals. It achieves the technical effect of improving the accuracy of terminal anomaly detection and the reliability of fault tracing through adaptive backtracking feature analysis and multi-degree-of-freedom causal tracing, and realizing efficient and intelligent remote monitoring. If the collected anomaly depth coefficient is less than the collected anomaly depth threshold, multi-dimensional correlation latent fault inference is performed on the terminal monitoring multi-dimensional sequence based on the backtracking feature time zone to obtain the terminal latent fault inference map. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a remote monitoring method for an electricity information collection terminal device according to the present invention.

[0011] Figure 2 This is a schematic diagram of the structure of a remote monitoring system for an electricity information collection terminal device according to the present invention.

[0012] Figure labeling: Data acquisition module 11, data backtracking module 12, acquisition anomaly detection module 13, cause-effect tracing module 14, fault inference module 15. Detailed Implementation

[0013] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0014] Example 1, as Figure 1 This is a flowchart illustrating a remote monitoring method for an electricity consumption information collection terminal device according to the present invention. The remote monitoring method for the electricity consumption information collection terminal device includes:

[0015] Remotely monitor electricity consumption information collection terminal equipment to obtain terminal collection data and terminal monitoring multi-dimensional sequence corresponding to the current time node.

[0016] Specifically, during remote monitoring, the system first accesses the electricity information collection terminals deployed on the power user side or distribution equipment side in real time. Data synchronization is maintained with the terminals via the communication network to obtain all operational data collected by the terminals at the current time point. The obtained terminal data typically includes power parameters such as voltage, current, power, active power, reactive power, and power factor. This data comprehensively reflects the real-time collection status of the terminals at that moment. Simultaneously, the system extracts and organizes multi-dimensional monitoring sequences related to the terminal's operational status from the terminal equipment. These multi-dimensional terminal monitoring sequences include monitoring indicators from various dimensions such as hardware monitoring sequences, software monitoring sequences, and communication monitoring sequences. By simultaneously acquiring the terminal's collected data and the multi-dimensional terminal monitoring sequences at the current moment, the system can comprehensively grasp the real-time collection behavior and status characteristics of the collection terminal, providing basic data support for subsequent data backtracking, anomaly detection, and causal analysis, ensuring the real-time nature, accuracy, and traceability of the monitoring process.

[0017] In some embodiments, the terminal monitoring multidimensional sequence includes a terminal hardware monitoring sequence, a terminal software monitoring sequence, a terminal communication monitoring sequence, and a terminal environment monitoring sequence.

[0018] Specifically, the multi-dimensional monitoring sequence of the terminal is a set of monitoring data that comprehensively depicts the operating status of the electricity information collection terminal equipment. It reflects the dynamic changes in the terminal's hardware, software, communication, and environment. The terminal hardware monitoring sequence records indicators related to the operation of the terminal's physical hardware, such as processor load, memory usage, motherboard temperature, power input stability, battery level, and the operating current and voltage of key components. This information is used to determine potential risks such as hardware aging, overload operation, or abnormal temperature rise. The terminal software monitoring sequence characterizes the operating status of the terminal's internal software system, including operating system task scheduling and various service modules. The monitoring system includes four types of sequences: Execution status of blocks, program exception logs, cache overflow status, and thread blocking status, used to identify software logic anomalies, failures, or decreased operational efficiency; Terminal communication monitoring sequences, used to record the communication quality between the terminal and the master station or relay equipment, including communication link stability, message transmission and reception success rate, communication latency, data retransmission rate, network interference intensity, and abnormal disconnection records, used to assess whether the communication channel has attenuation, faults, or external interference; Terminal environment monitoring sequences, used to reflect the external working environment conditions of the terminal, such as ambient temperature, humidity, electromagnetic interference level, vibration, and power supply environment changes, used to determine whether the external environment has an adverse impact on terminal operation. Through continuous collection and dynamic updating of these four types of monitoring sequences, a multi-dimensional monitoring data foundation covering the entire lifecycle of terminal operation can be formed, providing comprehensive, accurate, and traceable input data for subsequent retrospective analysis, anomaly detection, causal analysis, and latent fault deduction, ensuring the scientific validity and reliability of the monitoring methods.

[0019] Based on the multidimensional sequence monitored by the terminal, adaptive backtracking feature analysis is performed on the current time node to construct a backtracking feature time zone. Then, based on the backtracking feature time zone, the electricity information collection terminal device is used to backtrack the collected data to obtain a backtracking collection dataset.

[0020] Specifically, after acquiring the multi-dimensional monitoring sequence of the terminal at the current time point, the system first performs adaptive backtracking feature analysis on the current time point based on the change patterns of multi-dimensional monitoring data such as terminal hardware, software, communication, and environment. By analyzing the baseline deviation of each monitoring dimension, the system dynamically determines the length of the backtracking window that best reflects the terminal's operating status. The resulting backtracking feature time zone can automatically adapt to the terminal's current operating status. When the monitoring sequence fluctuates significantly, the backtracking time zone increases accordingly; when the sequence is stable, the backtracking time zone shrinks, ensuring that the backtracking range is neither too large (causing redundancy) nor too small (causing information loss). After constructing the backtracking feature time zone, the system performs data backtracking on the electricity information collection terminal equipment from historical collection records according to the time range defined by this time zone, extracting all terminal collection data within the corresponding time period, and finally forming a backtracking collection dataset for subsequent anomaly detection and analysis. Through this method, the backtracking dataset can accurately cover key historical segments that affect the current state, improving the accuracy and interpretability of subsequent anomaly analysis.

[0021] In some embodiments, adaptive backtracking feature parsing is performed on the current time node based on the terminal monitoring multidimensional sequence to construct a backtracking feature time zone. The method includes:

[0022] A baseline deviation evaluation is performed based on the multidimensional sequence monitored by the terminal to obtain a terminal baseline deviation sequence. A parameter-by-parameter deviation sensitivity calculation is then performed on the terminal baseline deviation sequence based on a baseline deviation threshold to obtain multiple baseline deviation sensitivity coefficients. The mean of these multiple baseline deviation sensitivity coefficients is calculated to obtain a terminal baseline deviation sensitivity coefficient. A predetermined backtracking window is adaptively adjusted based on the terminal baseline deviation sensitivity coefficient to obtain an adjusted backtracking window. The backtracking time zone is configured for the current time node based on the adjusted backtracking window to generate the backtracking feature time zone.

[0023] Specifically, after acquiring the multi-dimensional monitoring sequence of the terminal, the system first evaluates the benchmark deviation of the operating parameters of each monitoring dimension. In this process, the system treats each monitoring parameter in the hardware, software, communication, and environmental monitoring sequences as an independent evaluation object, calculating its deviation from the corresponding preset operating benchmark and taking its absolute value. This absolute value is then divided by the corresponding preset operating benchmark to obtain multiple benchmark deviation evaluation coefficients. These evaluation coefficients are then combined in chronological order or by monitoring parameter order to form the terminal benchmark deviation sequence. Subsequently, based on a preset benchmark deviation threshold, the system performs parameter-by-parameter deviation sensitivity calculation on each benchmark deviation evaluation coefficient in the terminal benchmark deviation sequence. That is, by dividing the benchmark deviation evaluation coefficient of each parameter by the corresponding benchmark deviation threshold, the system calculates the benchmark deviation sensitivity coefficient for each parameter. Finally, the average of these benchmark deviation sensitivity coefficients is calculated to obtain the terminal benchmark deviation sensitivity coefficient that comprehensively reflects the overall deviation of the terminal. Subsequently, the system adaptively adjusts the originally set backtracking window, allowing the backtracking time range to automatically expand or shrink according to the current state of the terminal. During this process, the system adds the product of the predetermined backtracking window and the terminal's baseline deviation sensitivity coefficient to calculate the required adjusted backtracking window. When the terminal's baseline deviation sensitivity coefficient is large, the adjusted backtracking window is correspondingly larger to cover a wider historical time period; when the deviation is small, the backtracking window remains small to avoid introducing too much irrelevant data. Finally, the system configures the backtracking time zone for the current time node based on the adjusted backtracking window, generating a backtracking feature time zone that adapts to the current operating characteristics, providing a precise data foundation for subsequent data backtracking and anomaly detection.

[0024] Based on the backtracking dataset, the terminal-collected data is subjected to data acquisition anomaly depth detection to obtain the data acquisition anomaly depth coefficient.

[0025] Specifically, after constructing the backtracking data collection dataset, ARIMA (Autoregressive Integral Moving Average) and LSTM (Long Short-Term Memory) are used to perform data collection anomaly depth detection on the terminal data collected at the current time point. This quantifies the deviation between the current collection behavior and historical collection patterns, resulting in a first evaluation coefficient for collection anomaly prediction by ARIMA and a second evaluation coefficient for collection anomaly prediction by LSTM. Then, these two evaluation coefficients are fused with equal weights to generate a data collection anomaly depth coefficient, which visually reflects the difference between the current collection behavior and historical normal behavior. A larger data collection anomaly depth coefficient indicates a more significant deviation between the current collection state and historical stable patterns; a smaller coefficient indicates that the current collection state is within the normal fluctuation range, thus providing a quantifiable basis for subsequent determination of whether to perform causal tracing or latent fault inference.

[0026] In some embodiments, the method includes performing acquisition anomaly depth detection on the terminal acquisition data based on the backtracking acquisition dataset to obtain an acquisition anomaly depth coefficient.

[0027] Based on the backtracking data collection dataset, ARIMA prediction of terminal acquisition features is performed on the current time node to obtain a first acquisition feature prediction result; based on the backtracking data collection dataset, LSTM prediction of terminal acquisition features is performed on the current time node to obtain a second acquisition feature prediction result; based on the first acquisition feature prediction result, acquisition variation depth evaluation is performed on the terminal acquisition data to obtain a first acquisition variation evaluation coefficient; based on the second acquisition feature prediction result, acquisition variation depth evaluation is performed on the terminal acquisition data to obtain a second acquisition variation evaluation coefficient; based on the first acquisition variation evaluation coefficient and the second acquisition variation evaluation coefficient, equal weighted fusion is performed to generate the acquisition variation depth coefficient.

[0028] Specifically, after acquiring the retrospective data collection dataset, the system first performs dual-model prediction on the terminal collection behavior at the current time point based on the multi-period historical collection features contained in the dataset, in order to improve the stability and accuracy of the prediction results. In this process, the system inputs the retrospective data collection dataset into a pre-built ARIMA time series prediction model, and obtains the first collection feature prediction result for the current time point by fitting the trend term, period term, and random disturbance term of the historical collection data, which is used to reflect the expected collection value under linear time series patterns. At the same time, the retrospective data collection dataset is input into a pre-built LSTM deep learning model, and the system performs deep prediction on the collection features at the current time point by capturing the long-term dependencies and nonlinear dynamic change trends in the collection data, generating the second collection feature prediction result. The ARIMA time series prediction model is constructed through parameter modeling and fitting of autoregressive, differencing, and moving average terms; the LSTM deep learning model is constructed through forward propagation, loss calculation, backpropagation, parameter optimization, and other steps. After obtaining the two types of prediction results mentioned above, the system uses the first and second acquisition feature prediction results as benchmarks to evaluate the depth of acquisition variation of the actual terminal acquisition data at the current time point. Specifically, the system calculates the absolute difference between the actual acquisition value and the ARIMA predicted value in the first acquisition feature prediction result, divides the calculated difference by the ARIMA predicted value to obtain the relative deviation rate, and then divides the calculated difference by the historical standard deviation of the corresponding acquisition feature in the retrospective acquisition dataset to obtain the historical fluctuation normalized deviation. The first evaluation coefficient of acquisition variation is obtained by averaging the relative deviation rate and the historical fluctuation normalized deviation. Simultaneously, the second evaluation coefficient of acquisition variation between the actual acquisition value and the LSTM predicted value is calculated using the same method. These two evaluation coefficients reflect the degree of anomaly in acquisition behavior under different prediction model benchmarks. Subsequently, in order to obtain a more comprehensive and robust assessment of the degree of variation in the current acquisition status, the system performs equal-weighted fusion of the first and second evaluation coefficients of acquisition variation, and then averages and combines them to finally generate a unified acquisition variation depth coefficient that characterizes the degree of anomaly in the current acquisition features. This acquisition variation depth coefficient also incorporates deviation information from linear time-series prediction and nonlinear depth prediction, which can more accurately reflect the difference between the current acquisition behavior and historical stable patterns, providing a reliable quantitative basis for subsequent causal tracing or latent fault inference.

[0029] If the collected mutation depth coefficient is greater than or equal to the collected mutation depth threshold, multi-degree-of-freedom causal tracing is performed on the collected mutation depth coefficient according to the terminal monitoring multi-dimensional sequence to obtain the terminal mutation causal tracing map.

[0030] Specifically, when the acquisition anomaly depth coefficient calculated by the system is greater than or equal to the preset acquisition anomaly depth threshold, it indicates that the current terminal acquisition characteristics have shown significant anomalies, and the system needs to further determine the root cause of the anomaly. To this end, the system activates a multi-degree-of-freedom causal tracing mechanism, using the acquisition anomaly depth coefficient as the starting point. This mechanism combines multi-dimensional sequences from terminal monitoring to perform multi-path, multi-dimensional causal tracing of anomalies, gradually constructing a path from the abnormal phenomenon to the intermediate triggering factor, and then to the potential root cause. After the causal paths formed by each monitoring dimension are independently recorded, the system structurally integrates all paths to generate a terminal anomaly causal tracing map covering all dimensions of hardware, software, communication, and environment. This terminal anomaly causal tracing map intuitively presents the multi-source triggering links and potential root causes of acquisition anomalies, providing a clear basis for subsequent anomaly localization and fault handling.

[0031] In some embodiments, the method includes performing multi-degree-of-freedom causal tracing of the collected mutation depth coefficients based on the terminal monitoring multidimensional sequence to obtain a terminal mutation causal tracing map.

[0032] The collected mutation depth coefficients are traced causally according to the terminal hardware monitoring sequence to obtain a first causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal software monitoring sequence to obtain a second causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal communication monitoring sequence to obtain a third causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal environment monitoring sequence to obtain a fourth causal tracing path; the first, second, third, and fourth causal tracing paths are combined to generate the terminal mutation causal tracing map.

[0033] Specifically, after the collected anomaly depth coefficient reaches the anomaly trigger threshold, the system initiates a multi-dimensional causal tracing process. Based on the hardware monitoring sequence, software monitoring sequence, communication monitoring sequence, and environmental monitoring sequence in the terminal monitoring multi-dimensional sequence, the system performs a multi-dimensional analysis of the source of the collected anomaly. First, the system uses the collected anomaly depth coefficient as the starting point of the causal analysis, treating it as the top-level anomaly node. Based on the hardware operating parameters related to power status, chip temperature, memory health, and the operating current of the collected module in the terminal hardware monitoring sequence, the system analyzes which hardware features might trigger the anomaly, thus forming a correlation coefficient link from the top-level anomaly to the intermediate hardware triggering features, and then to the bottom-level hardware root cause. The first path of anomaly causal tracing is constructed through Boolean logic association. Similarly, based on the terminal software monitoring sequence, terminal communication monitoring sequence, and terminal environment monitoring sequence, the system uses the corresponding acquisition anomaly depth coefficient as the starting point for causal analysis. Based on the corresponding software operating parameters, communication parameters, and environmental parameters, it analyzes which features might trigger the anomaly, thus forming a correlation coefficient link from the top-level anomaly to the intermediate triggering features, and then to the bottom-level root cause. Through Boolean logic association, it constructs a second, third, and fourth path for anomaly causal tracing. After obtaining the four causal paths, the system uniformly organizes all paths, structuring and visually representing the top-level anomaly nodes, intermediate triggering nodes, and bottom-level root cause nodes contained in each path according to the causal chain relationship. The final generated terminal anomaly causal tracing map can simultaneously present the anomaly causes and triggering logic under multi-dimensional monitoring sequences, forming a multi-degree-of-freedom causal explanation structure across hardware, software, communication, and environmental dimensions, providing complete data support and logical evidence for accurately locating the root cause of the acquisition anomaly.

[0034] In some embodiments, the method includes performing causal tracing of the collected mutation depth coefficients based on the terminal hardware monitoring sequence to obtain a first causal tracing path for mutation, the method comprising:

[0035] Using the collected mutation depth coefficient as the top-level mutation node; tracing the intermediate trigger features of the top-level mutation node according to the terminal hardware monitoring sequence to obtain the intermediate association tracing result of the mutation hardware; tracing the basic trigger features of the intermediate association tracing result of the mutation hardware according to the terminal hardware monitoring sequence to obtain the basic association tracing result of the mutation hardware; performing Boolean logic association based on the top-level mutation node, the intermediate association tracing result of the mutation hardware, and the basic association tracing result of the mutation hardware to generate the first causal tracing path of the mutation.

[0036] Specifically, in the causal tracing process at the hardware level, the system first uses the acquisition anomaly depth coefficient as the starting point node for causal analysis, defining it as the top-level anomaly node to represent the overall degree of anomaly in the current terminal acquisition behavior. Subsequently, based on the terminal hardware monitoring sequence, the system performs intermediate trigger feature tracing on this top-level anomaly node. During this process, the system selects mid-level hardware operating parameters from the hardware monitoring sequence that may be correlated with the acquisition anomaly, including but not limited to temperature changes in the acquisition chip, voltage stability of the power supply module, workload of the sensor interface, noise level of the current sampling circuit, internal bus status, and memory access latency. It then analyzes the fluctuation trends, deviation magnitudes, and synchronicity of these hardware parameters before and after the anomaly. Through statistical correlation analysis or rule matching, it determines the intermediate trigger features associated with the acquisition anomaly depth coefficient, forming the intermediate correlation tracing results for the anomaly hardware. For example, the Pearson correlation coefficient can be used to measure the correlation between each hardware operating parameter and the acquisition anomaly depth coefficient, and compared with a set hardware correlation threshold. When the correlation of a hardware operating parameter exceeds the correlation threshold, the system marks that hardware operating parameter as an intermediate hardware feature that may trigger the acquisition anomaly. After obtaining the intermediate correlation tracing results of the abnormal hardware, the system further delves into the intermediate hardware characteristics within these results, analyzing the underlying hardware factors that trigger these intermediate characteristics. These factors include the original input voltage of the power supply module, peak power supply ripple, temperature of key internal hardware nodes, operating current of the main control chip, base noise of the sensor channel, electromagnetic interference intensity, line impedance changes, and power load fluctuations—all potential fundamental triggering characteristics. Subsequently, the system uses the intermediate triggering characteristics as the target variable and the fundamental triggering characteristics as candidate root cause variables. It employs a conditional correlation coefficient calculation method to statistically derive the relationship between the two. The conditional correlation coefficient can be calculated using partial correlation coefficients, or it can be calculated using conditional correlation coefficients or Granger causality test statistical indicators. By calculating the partial correlation coefficients, it can determine whether the fundamental triggering characteristics still significantly affect the intermediate triggering characteristics after controlling and acquiring abnormal influencing factors. If the partial correlation coefficient reaches a set threshold, the system considers the fundamental triggering characteristics to constitute a potential root cause of the intermediate triggering characteristics. Then, the system repeats the above calculation for all basic trigger features and sorts all basic trigger features that meet the conditions according to the absolute value of their partial correlation coefficients to form the basic correlation tracing result of the hardware dimension. This basic correlation tracing result of the abnormal hardware represents the lowest-level hardware cause that drives the change of intermediate trigger features and can be used to construct a complete causal chain from abnormal phenomenon → intermediate trigger feature → root cause feature.Ultimately, the system performs Boolean logic associations on the top-level nodes of the mutation, the intermediate correlation results obtained from tracing intermediate trigger features, and the bottom-level correlation results obtained from tracing basic trigger features, according to the causal hierarchy. That is, by establishing a logical structure of "AND," "OR," and "NOT" relationships, the system combines each node into a complete hardware-dimensional causal path, enabling the path to reflect the hierarchical causal relationship from top-level anomaly to intermediate triggering factors to bottom-level hardware root causes. The logically organized structure constitutes the first path of mutation causal tracing, providing a complete hardware-level causal chain basis for the subsequent construction of a multi-dimensional causal graph.

[0037] If the acquisition variation depth coefficient is less than the acquisition variation depth threshold, multidimensional correlation latent fault inference is performed on the terminal monitoring multidimensional sequence based on the retrospective characteristic time zone to obtain the terminal latent fault inference map.

[0038] Specifically, when the acquisition anomaly depth coefficient is less than the acquisition anomaly depth threshold, the system considers that the current terminal acquisition status has not produced significant anomalies, but may still hide potential minor faults or early degradation signs. To this end, based on the previously constructed retrospective feature time zone, the system performs multi-dimensional correlation latent fault inference on the multi-dimensional sequence of terminal monitoring. That is, using the time range defined by the retrospective feature time zone, multi-dimensional retrospective analysis is performed on the terminal hardware monitoring sequence, terminal software monitoring sequence, terminal communication monitoring sequence, and terminal environment monitoring sequence to extract the key feature changes of each monitoring dimension within the time zone. Then, based on these features, latent fault inference analysis is performed to determine whether it conforms to the evolution pattern of known minor hardware anomalies, minor software misalignments, communication fluctuation trends, or early characteristics of environmental disturbances, and latent fault inference results are generated for the four dimensions of hardware, software, communication, and environment, respectively. Finally, the system integrates the latent fault projection results from the above dimensions in a structured manner to form a terminal latent fault projection map that includes multi-dimensional potential fault links, mild abnormality characteristics and evolution possibilities. This map is used to present the potential fault risks and future evolution direction of the terminal in the absence of significant abnormalities, thereby providing a reference for early warning and maintenance of the equipment.

[0039] In some embodiments, a multidimensional correlation latent fault inference is performed on the terminal monitoring multidimensional sequence based on the retrospective characteristic time zone to obtain a terminal latent fault inference map. The method includes:

[0040] Based on the backtracking characteristic time zone, multi-dimensional monitoring backtracking is performed on the electricity information collection terminal equipment to obtain hardware monitoring backtracking sets, software monitoring backtracking sets, communication monitoring backtracking sets, and environmental monitoring backtracking sets. Based on the hardware monitoring backtracking sets, latent faults are inferred from the terminal hardware monitoring sequences to obtain hardware-related latent fault inference results. Based on the software monitoring backtracking sets, latent faults are inferred from the terminal software monitoring sequences to obtain software-related latent fault inference results. Based on the communication monitoring backtracking sets, latent faults are inferred from the terminal communication monitoring sequences to obtain communication-related latent fault inference results. Based on the environmental monitoring backtracking sets, latent faults are inferred from the terminal environmental monitoring sequences to obtain environmental-related latent fault inference results. The hardware-related latent fault inference results, the software-related latent fault inference results, the communication-related latent fault inference results, and the environmental-related latent fault inference results are analyzed to generate the terminal latent fault inference map.

[0041] Specifically, after determining that the collected anomaly depth coefficient has not reached the anomaly trigger threshold, in order to identify minor faults that may be in their early stages, the system extracts hardware monitoring backtracking sets, software monitoring backtracking sets, communication monitoring backtracking sets, and environmental monitoring backtracking sets from the multi-dimensional terminal monitoring sequences according to the time range defined by the backtracking feature time zone. These backtracking sets reflect the fine-grained dynamic changes of each monitoring dimension within that feature time zone. Subsequently, the system performs latent fault inference on the corresponding monitoring sequences based on the aforementioned backtracking sets. In the hardware dimension, the system performs feature analysis on hardware parameters based on the hardware monitoring backtracking sets, extracting characteristic changes that can reflect the trend of minor hardware anomalies, and uses these as input features for hardware latent fault inference to obtain hardware-related latent fault inference results. In the software dimension, the system extracts features from software operating state parameters based on the software monitoring backtracking sets, identifies minor fluctuation characteristics at the software level, and uses these characteristics in the software latent fault inference channel to generate software-related latent fault inference results. In the communication dimension, the system analyzes changes in communication link characteristics based on the communication monitoring backtracking set, extracts slightly correlated features of potential communication anomalies, and constructs inputs for communication latent fault inference to obtain communication-related latent fault inference results. In the environmental dimension, the system extracts feature changes related to environmental disturbances through the environmental monitoring backtracking set, uses these slight shifts in environmental features as the basis for inference, and forms environmental-related latent fault inference results. After obtaining the latent fault inference results in the four dimensions, the system uniformly sorts and structurally integrates these results, and constructs a multi-dimensional, multi-path latent fault inference map based on the potential coupling relationships and common triggering modes between fault features in each dimension. The final terminal latent fault inference map can comprehensively display the potential fault risks, possible evolution directions, and cross-dimensional impact links of power consumption information collection terminal equipment in the stage of slight anomalies, thereby providing reliable data support for early warning, maintenance decisions, and operation optimization.

[0042] In some embodiments, the method involves performing latent fault deduction on the terminal hardware monitoring sequence based on the hardware monitoring backtracking set to obtain hardware-related latent fault deduction results, the method comprising:

[0043] Based on the hardware monitoring backtracking set, mutation analysis is performed on the terminal hardware monitoring sequence to construct a terminal hardware characteristic mutation vector; a hardware-related terminal fault log set is loaded, where each hardware-related terminal fault log includes hardware characteristic mutation samples and terminal fault samples corresponding to the electricity consumption information collection terminal device; the imbalance of fault sample distribution is evaluated based on the hardware-related terminal fault log set to obtain the sample distribution imbalance degree; based on the sample distribution imbalance degree, an adversarial sample generator is used to perform perturbation optimization on the hardware-related terminal fault log set to establish a hardware-related fault sample optimization space; the hardware-related fault sample optimization space is adaptively and dynamically tuned and trained based on the AdamW optimizer to establish a hardware-related latent fault inference channel; the terminal hardware characteristic mutation vector is input into the hardware-related latent fault inference channel to output the hardware-related latent fault inference result.

[0044] Specifically, after acquiring the hardware monitoring backtracking set, the system first performs mutation analysis on the hardware operating parameters in the set. This involves a refined analysis of changes in hardware characteristics such as power supply voltage, chip temperature, memory access characteristics, sensor interface response time, and noise level of the acquisition circuit. Data points whose changes exceed thresholds within a short period are identified as parameter mutation points, and their amplitude and rate of change are extracted. These features are then combined to form a terminal hardware characteristic mutation vector, used to characterize minor hardware anomalies within the backtracking time zone. Subsequently, the system loads a pre-built hardware-related terminal fault log set. This log set contains a large amount of historical fault information accumulated during actual terminal operation. Each log entry includes a hardware characteristic mutation sample and a corresponding terminal fault sample, describing which mutation mode might correspond to which fault. The system then performs statistical analysis on the quantity distribution of various fault samples in the log set. By calculating the difference in sample proportions between different fault categories, the overall sample distribution imbalance is obtained, used to measure the degree of deviation between different fault categories in the log set. Subsequently, the system controls the adversarial example generator (built based on a Generative Adversarial Network, GAN) based on the imbalance in sample distribution, perturbing and optimizing the original samples in the hardware-related terminal fault log set to generate more fitted samples for scarce categories, thereby enhancing the expressive power of small and weak sample categories. After perturbation optimization, the system constructs a hardware-related fault sample optimization space that covers a more comprehensive and balanced feature distribution. Then, the system adaptively and dynamically tunes this optimization space based on the AdamW optimizer. During training, AdamW can adaptively adjust the learning rate according to gradient changes and suppress overfitting through weight decay, enabling the inference model built based on deep neural networks to more stably learn the implicit mapping relationship between hardware mutation features and fault risk from the hardware-related fault sample optimization space. After training, the system establishes a hardware-related implicit fault inference channel. Finally, when the terminal hardware characteristic mutation vector is input into the latent fault inference channel, the system outputs the corresponding hardware-related latent fault inference result based on the pre-trained mapping rules in the hardware-related latent fault inference channel. This hardware-related latent fault inference result can reflect the potential latent fault types and their occurrence probabilities that may correspond to the current hardware mutation characteristics, providing hardware-dimensional inference basis for the multidimensional latent fault inference map.

[0045] In summary, the remote monitoring method for an electricity information collection terminal device provided by this invention has the following technical effects:

[0046] A remote monitoring terminal device for electricity consumption information collection obtains terminal collection data and a multi-dimensional sequence of terminal monitoring corresponding to the current time node. Based on the multi-dimensional sequence of terminal monitoring, adaptive backtracking feature analysis is performed on the current time node to construct a backtracking feature time zone. The collected data from the terminal device is then backtracked based on the backtracking feature time zone to obtain a backtracking collection dataset. Based on the backtracking collection dataset, the collected data undergoes depth detection of collection anomalies to obtain a collection anomaly depth coefficient. If the collection anomaly depth coefficient is greater than or equal to a collection anomaly depth threshold, multi-degree-of-freedom causal tracing is performed on the collection anomaly depth coefficient based on the multi-dimensional sequence of terminal monitoring to obtain a terminal anomaly causal tracing map. If the collection anomaly depth coefficient is less than the collection anomaly depth threshold, multi-dimensional correlation latent fault inference is performed on the multi-dimensional sequence of terminal monitoring based on the backtracking feature time zone to obtain a terminal latent fault inference map. This achieves the technical effect of improving the accuracy of terminal anomaly detection and the reliability of fault tracing through adaptive backtracking feature analysis and multi-degree-of-freedom causal tracing, thus realizing efficient and intelligent remote monitoring.

[0047] Example 2, as Figure 2 This is a schematic diagram of the structure of a remote monitoring system for an electricity information collection terminal device according to the present invention. For example, Figure 1 A flowchart illustrating a remote monitoring method for an electricity information collection terminal device according to the present invention can be used as follows: Figure 2 The structure shown is implemented.

[0048] Based on the same concept as the remote monitoring method for an electricity information collection terminal device described in the above embodiment, the present invention also provides a remote monitoring system for an electricity information collection terminal device, comprising:

[0049] Data Acquisition Module 11: Remotely monitors the electricity consumption information acquisition terminal equipment to obtain the terminal acquisition data and terminal monitoring multidimensional sequence corresponding to the current time node; Data Backtracking Module 12: Performs adaptive backtracking feature analysis on the current time node based on the terminal monitoring multidimensional sequence, constructs a backtracking feature time zone, and performs backtracking of the acquisition data of the electricity consumption information acquisition terminal equipment based on the backtracking feature time zone to obtain a backtracking acquisition dataset; Acquisition Anomaly Detection Module 13: Performs acquisition anomaly depth detection on the terminal acquisition data based on the backtracking acquisition dataset to obtain an acquisition anomaly depth coefficient; Causal Tracing Module 14: If the acquisition anomaly depth coefficient is greater than or equal to the acquisition anomaly depth threshold, performs multi-degree-of-freedom causal tracing on the acquisition anomaly depth coefficient based on the terminal monitoring multidimensional sequence to obtain a terminal anomaly causal tracing map; Fault Inference Module 15: If the acquisition anomaly depth coefficient is less than the acquisition anomaly depth threshold, performs multi-dimensional correlation latent fault inference on the terminal monitoring multidimensional sequence based on the backtracking feature time zone to obtain a terminal latent fault inference map.

[0050] In some embodiments, the data acquisition module 11 includes:

[0051] The terminal monitoring multidimensional sequence includes a terminal hardware monitoring sequence, a terminal software monitoring sequence, a terminal communication monitoring sequence, and a terminal environment monitoring sequence.

[0052] In some embodiments, the data backtracking module 12 includes:

[0053] A baseline deviation evaluation is performed based on the multidimensional sequence monitored by the terminal to obtain a terminal baseline deviation sequence. A parameter-by-parameter deviation sensitivity calculation is then performed on the terminal baseline deviation sequence based on a baseline deviation threshold to obtain multiple baseline deviation sensitivity coefficients. The mean of these multiple baseline deviation sensitivity coefficients is calculated to obtain a terminal baseline deviation sensitivity coefficient. A predetermined backtracking window is adaptively adjusted based on the terminal baseline deviation sensitivity coefficient to obtain an adjusted backtracking window. The backtracking time zone is configured for the current time node based on the adjusted backtracking window to generate the backtracking feature time zone.

[0054] In some embodiments, the mutation detection module 13 includes:

[0055] Based on the backtracking data collection dataset, ARIMA prediction of terminal acquisition features is performed on the current time node to obtain a first acquisition feature prediction result; based on the backtracking data collection dataset, LSTM prediction of terminal acquisition features is performed on the current time node to obtain a second acquisition feature prediction result; based on the first acquisition feature prediction result, acquisition variation depth evaluation is performed on the terminal acquisition data to obtain a first acquisition variation evaluation coefficient; based on the second acquisition feature prediction result, acquisition variation depth evaluation is performed on the terminal acquisition data to obtain a second acquisition variation evaluation coefficient; based on the first acquisition variation evaluation coefficient and the second acquisition variation evaluation coefficient, equal weighted fusion is performed to generate the acquisition variation depth coefficient.

[0056] In some embodiments, the causal tracing module 14 includes:

[0057] The collected mutation depth coefficients are traced causally according to the terminal hardware monitoring sequence to obtain a first causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal software monitoring sequence to obtain a second causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal communication monitoring sequence to obtain a third causal tracing path; the collected mutation depth coefficients are traced causally according to the terminal environment monitoring sequence to obtain a fourth causal tracing path; the first, second, third, and fourth causal tracing paths are combined to generate the terminal mutation causal tracing map.

[0058] In some embodiments, the causal tracing module 14 includes:

[0059] Using the collected mutation depth coefficient as the top-level mutation node; tracing the intermediate trigger features of the top-level mutation node according to the terminal hardware monitoring sequence to obtain the intermediate association tracing result of the mutation hardware; tracing the basic trigger features of the intermediate association tracing result of the mutation hardware according to the terminal hardware monitoring sequence to obtain the basic association tracing result of the mutation hardware; performing Boolean logic association based on the top-level mutation node, the intermediate association tracing result of the mutation hardware, and the basic association tracing result of the mutation hardware to generate the first causal tracing path of the mutation.

[0060] In some embodiments, the fault deduction module 15 includes:

[0061] Based on the backtracking characteristic time zone, multi-dimensional monitoring backtracking is performed on the electricity information collection terminal equipment to obtain hardware monitoring backtracking sets, software monitoring backtracking sets, communication monitoring backtracking sets, and environmental monitoring backtracking sets. Based on the hardware monitoring backtracking sets, latent faults are inferred from the terminal hardware monitoring sequences to obtain hardware-related latent fault inference results. Based on the software monitoring backtracking sets, latent faults are inferred from the terminal software monitoring sequences to obtain software-related latent fault inference results. Based on the communication monitoring backtracking sets, latent faults are inferred from the terminal communication monitoring sequences to obtain communication-related latent fault inference results. Based on the environmental monitoring backtracking sets, latent faults are inferred from the terminal environmental monitoring sequences to obtain environmental-related latent fault inference results. The hardware-related latent fault inference results, the software-related latent fault inference results, the communication-related latent fault inference results, and the environmental-related latent fault inference results are analyzed to generate the terminal latent fault inference map.

[0062] In some embodiments, the fault deduction module 15 includes:

[0063] Based on the hardware monitoring backtracking set, mutation analysis is performed on the terminal hardware monitoring sequence to construct a terminal hardware characteristic mutation vector; a hardware-related terminal fault log set is loaded, where each hardware-related terminal fault log includes hardware characteristic mutation samples and terminal fault samples corresponding to the electricity consumption information collection terminal device; the imbalance of fault sample distribution is evaluated based on the hardware-related terminal fault log set to obtain the sample distribution imbalance degree; based on the sample distribution imbalance degree, an adversarial sample generator is used to perform perturbation optimization on the hardware-related terminal fault log set to establish a hardware-related fault sample optimization space; the hardware-related fault sample optimization space is adaptively and dynamically tuned and trained based on the AdamW optimizer to establish a hardware-related latent fault inference channel; the terminal hardware characteristic mutation vector is input into the hardware-related latent fault inference channel to output the hardware-related latent fault inference result.

[0064] In embodiment three, the present invention also provides a computer-readable storage medium that can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the remote monitoring method of an electricity information collection terminal device in the embodiments of the present invention, thereby realizing the aforementioned remote monitoring method of an electricity information collection terminal device.

[0065] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications 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 included within the protection scope of this invention.

Claims

1. A remote monitoring method for an electricity information collection terminal device, characterized in that, The method includes: Remotely monitor electricity consumption information collection terminal equipment to obtain the terminal collection data and multi-dimensional terminal monitoring sequence corresponding to the current time node; Based on the multidimensional sequence monitored by the terminal, adaptive backtracking feature analysis is performed on the current time node to construct a backtracking feature time zone. Based on the backtracking feature time zone, the electricity information collection terminal device is used to backtrack the collected data to obtain a backtracking collection dataset. Based on the backtracking dataset, the terminal data is subjected to data acquisition anomaly depth detection to obtain the data acquisition anomaly depth coefficient. If the collected mutation depth coefficient is greater than or equal to the collected mutation depth threshold, multi-degree-of-freedom causal tracing is performed on the collected mutation depth coefficient according to the terminal monitoring multi-dimensional sequence to obtain the terminal mutation causal tracing map. If the acquisition variation depth coefficient is less than the acquisition variation depth threshold, multidimensional correlation latent fault inference is performed on the terminal monitoring multidimensional sequence based on the retrospective characteristic time zone to obtain the terminal latent fault inference map. Specifically, the acquisition anomaly depth detection is performed on the terminal acquisition data based on the backtracking acquisition dataset to obtain the acquisition anomaly depth coefficient, including: Based on the backtracking dataset, perform terminal acquisition feature ARIMA prediction on the current time node to obtain the first acquisition feature prediction result; Based on the backtracking dataset, perform terminal acquisition feature LSTM prediction on the current time node to obtain the second acquisition feature prediction result; Based on the prediction result of the first acquisition feature, the acquisition anomaly depth evaluation is performed on the terminal acquisition data to obtain the first acquisition anomaly evaluation coefficient. Based on the prediction result of the second acquisition feature, the acquisition anomaly depth evaluation is performed on the terminal acquisition data to obtain the second acquisition anomaly evaluation coefficient. The depth coefficient of the acquisition anomaly is generated by performing equal-weighted fusion of the first evaluation coefficient and the second evaluation coefficient of the acquisition anomaly.

2. The remote monitoring method for an electricity information collection terminal device as described in claim 1, characterized in that, Based on the multi-dimensional sequence monitored by the terminal, adaptive backtracking feature parsing is performed on the current time node to construct a backtracking feature time zone, including: Based on the multidimensional sequence of terminal monitoring, a benchmark deviation evaluation is performed to obtain the terminal benchmark deviation sequence; Based on the benchmark deviation threshold, the terminal benchmark deviation sequence is calculated on a parameter-by-parameter basis to obtain multiple benchmark deviation sensitivity coefficients; The terminal baseline deviation sensitivity coefficient is obtained by averaging the multiple baseline deviation sensitivity coefficients. The predetermined backtracking window is adaptively adjusted based on the terminal reference deviation sensitivity coefficient to obtain the adjusted backtracking window; The backtracking time zone is configured for the current time node according to the adjustment backtracking window, and the backtracking feature time zone is generated.

3. The remote monitoring method for an electricity information collection terminal device as described in claim 1, characterized in that, Based on the multidimensional sequence of terminal monitoring, multi-degree-of-freedom causal tracing is performed on the collected mutation depth coefficients to obtain a terminal mutation causal tracing map, including: Based on the terminal hardware monitoring sequence, the collected mutation depth coefficient is traced back to the causal relationship of the mutation to obtain the first path of mutation causal tracing. Based on the monitoring sequence of the terminal software, the collected mutation depth coefficient is traced back to the causal relationship of the mutation to obtain a second path for causal tracing of the mutation. Based on the terminal communication monitoring sequence, the collected mutation depth coefficient is traced back to its causal origin to obtain a third path for causal origin tracing of mutation. Based on the terminal environment monitoring sequence, the collected mutation depth coefficient is traced back to the causal relationship of the mutation to obtain the fourth path of mutation causal tracing. The first path of mutation causality tracing, the second path of mutation causality tracing, the third path of mutation causality tracing, and the fourth path of mutation causality tracing are organized to generate the terminal mutation causality tracing map.

4. The remote monitoring method for an electricity information collection terminal device as described in claim 3, characterized in that, Based on the terminal hardware monitoring sequence, the collected anomaly depth coefficient is traced back to determine the first path of anomaly causality tracing, including: The aforementioned mutation depth coefficient is used as the top-level node of the mutation; Based on the terminal hardware monitoring sequence, the intermediate triggering feature tracing of the abnormal top-level node is performed to obtain the intermediate association tracing result of the abnormal hardware. Based on the terminal hardware monitoring sequence, the basic trigger feature tracing is performed on the intermediate association tracing result of the abnormal hardware to obtain the basic association tracing result of the abnormal hardware. Boolean logic association is performed based on the top-level node of the mutation, the intermediate association tracing result of the mutation hardware, and the basic association tracing result of the mutation hardware to generate the first path of mutation causal tracing.

5. The remote monitoring method for an electricity information collection terminal device as described in claim 1, characterized in that, Based on the retrospective characteristic time zone, multidimensional correlation latent fault inference is performed on the multidimensional sequence of terminal monitoring to obtain a terminal latent fault inference map, including: Based on the backtracking characteristic time zone, the electricity consumption information collection terminal equipment is monitored and backtracked in multiple dimensions to obtain hardware monitoring backtracking set, software monitoring backtracking set, communication monitoring backtracking set and environmental monitoring backtracking set; Based on the hardware monitoring backtracking set, the terminal hardware monitoring sequence is used to perform latent fault inference to obtain the hardware-related latent fault inference results. Based on the software monitoring backtracking set, the terminal software monitoring sequence is used to perform latent fault inference to obtain the software-related latent fault inference results; Based on the communication monitoring backtracking set, the terminal communication monitoring sequence is used to perform latent fault inference to obtain the communication-related latent fault inference results; Based on the environmental monitoring backtracking set, the terminal environmental monitoring sequence is used to perform latent fault inference to obtain the environmental-related latent fault inference results; By analyzing the results of the hardware-related latent fault simulation, the software-related latent fault simulation, the communication-related latent fault simulation, and the environment-related latent fault simulation, a terminal latent fault simulation map is generated.

6. The remote monitoring method for an electricity information collection terminal device as described in claim 5, characterized in that, Based on the hardware monitoring backtracking set, latent faults are inferred from the terminal hardware monitoring sequence to obtain hardware-related latent fault inference results, including: Based on the hardware monitoring backtracking set, mutation analysis is performed on the terminal hardware monitoring sequence to construct a terminal hardware characteristic mutation vector. Load the hardware-associated terminal fault log set. Each hardware-associated terminal fault log includes hardware characteristic mutation samples and terminal fault samples corresponding to the electricity information collection terminal device. Based on the hardware-associated terminal fault log set, an evaluation of the uneven distribution of fault samples is performed to obtain the degree of uneven distribution of samples. Based on the sample distribution imbalance control, the adversarial sample generator performs perturbation optimization on the hardware-related terminal fault log set to establish a hardware-related fault sample optimization space. Based on the AdamW optimizer, the optimization space of the hardware-related fault samples is adaptively and dynamically tuned and trained to establish a hardware-related hidden fault inference channel. The terminal hardware characteristic mutation vector is input into the hardware-related latent fault inference channel, and the hardware-related latent fault inference result is output.

7. The remote monitoring method for an electricity information collection terminal device as described in claim 1, characterized in that, The terminal monitoring multidimensional sequence includes a terminal hardware monitoring sequence, a terminal software monitoring sequence, a terminal communication monitoring sequence, and a terminal environment monitoring sequence.

8. A remote monitoring system for an electricity information collection terminal device, characterized in that, A remote monitoring method for an electricity consumption information collection terminal device as described in any one of claims 1-7, the system comprising: Data acquisition module: Remotely monitors electricity consumption information acquisition terminal equipment to obtain terminal acquisition data and multi-dimensional terminal monitoring sequences corresponding to the current time node; Data backtracking module: Based on the multi-dimensional sequence monitored by the terminal, adaptive backtracking feature analysis is performed on the current time node to construct a backtracking feature time zone, and the collected data of the electricity information collection terminal device is backtracked according to the backtracking feature time zone to obtain the backtracking collection dataset; The data acquisition anomaly detection module performs data acquisition anomaly depth detection on the terminal acquisition data based on the backtracking acquisition dataset to obtain the data acquisition anomaly depth coefficient; Causal tracing module: If the collected mutation depth coefficient is greater than or equal to the collected mutation depth threshold, perform multi-degree-of-freedom causal tracing on the collected mutation depth coefficient according to the terminal monitoring multi-dimensional sequence to obtain the terminal mutation causal tracing map; Fault inference module: If the acquisition variation depth coefficient is less than the acquisition variation depth threshold, perform multi-dimensional correlation latent fault inference on the terminal monitoring multi-dimensional sequence based on the backtracking characteristic time zone to obtain the terminal latent fault inference map.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a remote monitoring method for an electricity information collection terminal device as described in any one of claims 1 to 7.

Images