Intelligent switch cabinet anti-explosion abnormality monitoring method and system

Abstract

The present application relates to the technical field of switch cabinet state monitoring, and particularly relates to an intelligent switch cabinet anti-explosion abnormality monitoring method and system, which comprises a data acquisition module, a data storage and management module, a correlation analysis module, a time sequence prediction module, a risk assessment and early warning module, a communication interface module and a man-machine interaction module. The present application realizes early prediction and accurate diagnosis of switch cabinet explosion risks by setting a monitoring architecture integrating multi-source sensing and intelligent analysis, quantifies the contribution weight of each cause parameter to the failure result through a dynamic correlation analysis algorithm, and introduces a time sequence prediction model to establish an advanced warning mechanism for parameter changes, which can identify potential failure development paths in advance by analyzing the change trend of cause parameters such as current and voltage when direct failure indicators such as temperature and gas concentration have not yet exceeded the standard, thereby changing the accident handling mode from post-repair to pre-prevention, and improving the active safety protection capability of switch cabinet operation.

Description

Technical Field

[0001] This invention relates to the field of switchgear condition monitoring technology, and in particular to an intelligent switchgear anti-fire and explosion anomaly monitoring method and system. Background Technology

[0002] As a key device in the power system for power distribution, control and protection, the reliability of switchgear is directly related to the safety and stability of the power supply network. Due to long-term operation in harsh environments with high current and high voltage, switchgear is prone to hidden dangers such as loose connections, insulation aging and increased contact resistance. These hidden dangers may lead to local overheating, insulation deterioration or even internal arcing faults, ultimately causing serious fire and explosion accidents. Traditional switchgear monitoring methods mainly rely on periodic manual inspections and simple temperature measurements. This method is not only inefficient, but also difficult to detect latent faults in time, and can no longer meet the needs of modern smart grids for real-time perception and early warning of equipment status.

[0003] With the development of sensing technology, existing technologies have begun to use online monitoring devices to monitor single or a few parameters such as local temperature and partial discharge in switchgear. While this approach represents progress from manual inspection to automatic monitoring, it largely remains at the stage of data acquisition and over-limit alarms, lacking in-depth analysis of the correlations between multiple parameters. Its main drawbacks are: firstly, monitoring systems typically have fixed alarm thresholds, triggering an alarm only when the measured value exceeds the threshold. This delayed alarm response fails to provide sufficient buffer time for fault handling. Secondly, different types of faults often exhibit different characteristic parameter changes in their early stages, and isolated parameter monitoring cannot accurately identify complex fault modes, easily leading to false alarms or missed alarms, thus affecting the accuracy of early warnings.

[0004] Therefore, we propose an intelligent method and system for monitoring fire and explosion prevention in switchgear. By constructing a monitoring system that integrates multi-source sensing and intelligent analysis, we establish a correlation analysis model with result data and cause data as the core, and introduce a dynamic optimization mechanism with self-learning capabilities to provide early warning at the fault initiation stage. At the same time, we continuously optimize the diagnostic accuracy through continuous learning, providing a brand-new intelligent solution for fire and explosion prevention in switchgear. Summary of the Invention

[0005] To achieve the above objectives, this invention proposes an intelligent method for monitoring fire and explosion anomalies in switchgear, comprising the following steps:

[0006] Step 1: Multi-source sensing and data aggregation, using multi-source sensors deployed at key nodes inside the switch cabinet for comprehensive status detection;

[0007] Step 2: Data association modeling and archiving steps, establish the association between result data and cause data, define parameters that directly characterize the failure results as result data, define potential cause parameters as cause data, and construct a cause data mapping table with result data as the core.

[0008] Step 3: Dynamic correlation analysis and weight allocation. By calculating the curve similarity between each causal data sequence and the result data sequence, the contribution weight of each causal parameter to the failure result is quantified.

[0009] Step 4: Time series modeling and prediction weight calculation. Time series analysis methods are used to determine the optimal time lag interval of each inducing factor data relative to the result data, and a prediction model is established.

[0010] Step 5: Real-time risk prediction and early warning. Based on the weighting model and the timeliness model, a comprehensive risk index is calculated to achieve tiered early warning.

[0011] Step 6: Misjudgment feedback and model self-correction. When a misjudged sample appears, it is collected as a negative sample case. The weight allocation and warning threshold are dynamically adjusted to achieve self-learning optimization of the model.

[0012] In one example, the third step calculates the importance of different causal data under the same result data. In the same result data, different causal data corresponding to different state result data are selected, and a circuit diagram is generated for curvature calculation. The dynamic correlation strength between causal and result is quantified by analyzing the morphological similarity of the data curves.

[0013] In one example, the time series analysis algorithm in the fourth step automatically finds the optimal time lag interval between different causal data and outcome data, and assigns prediction weights to these lagged causal data.

[0014] In one example, after detecting significant fluctuations in one or more triggering data in step six and issuing an early warning accordingly, if no expected changes in the corresponding result data are observed within the set extended timeframe, the system will determine that the early warning is a misjudgment.

[0015] In one example, if the number of misjudgments in step six exceeds a set threshold, the case is classified as a negative sample. In the model self-correction process, the weight of the specific triggering data that caused the misjudgment is reduced, and the weight of the second weight is increased. The case is judged again. If the second weight still results in a misjudgment, the data records in the negative sample case are compared with the data in the database when the fault occurred. The cases with the same misjudged data are identified, and the triggering data with different values ​​under the same misjudged data is identified. The triggering data is combined with the previous misjudged data for monitoring, and the judgment is updated.

[0016] In one example, an intelligent switchgear fire and explosion prevention anomaly monitoring system includes a data acquisition module, a data storage and management module, a correlation analysis module, a time series prediction module, a risk assessment and early warning module, a communication interface module, and a human-computer interaction module.

[0017] In one example, the data acquisition module consists of multiple types of sensor units, signal conditioning units, and edge computing nodes, while the data storage and management module constructs a hybrid storage architecture of time-series databases and relational databases.

[0018] In one example, the association analysis module implements the function of dynamic association weight calculation, uses the dynamic time warping algorithm to calculate the curve similarity between the causal data sequence and the result data sequence, outputs the association weight value in the 0-1 interval, and has a weight update unit.

[0019] In one example, the time series prediction module integrates an autoregressive integral moving average model and a long short-term memory neural network algorithm. The lag analysis engine automatically performs cross-correlation function calculations to determine the optimal time lag interval of each causal data relative to the result data and generates a lag parameter table. The weighted time extension unit performs time shifting on the historical causal data according to the lag interval and constructs a prediction feature vector.

[0020] In one example, the risk assessment and early warning module includes a risk index calculator and a graded early warning generator; the self-learning optimization module consists of a misjudgment identification unit and a negative sample processing unit; the communication interface module enables connection with the substation integrated automation system; and the human-machine interaction module is based on a Web architecture, providing real-time monitoring screens, trend curve displays, early warning information lists, model parameter configuration, and historical data query functions, and supports mobile access.

[0021] The intelligent switchgear fire and explosion prevention anomaly monitoring method and system proposed in this invention can bring the following beneficial effects:

[0022] 1. This invention achieves early prediction and accurate diagnosis of switchgear explosion risks by setting up a monitoring architecture that integrates multi-source sensing and intelligent analysis. It quantifies the contribution weight of each inducing parameter to the fault outcome through dynamic correlation analysis algorithm, and introduces a time-series prediction model to establish an early warning mechanism for parameter changes. It can identify potential fault development paths in advance by analyzing the changing trends of inducing parameters such as current and voltage before direct fault indicators such as temperature and gas concentration exceed the standard. This transforms the accident handling mode from post-event remediation to pre-event prevention, significantly improving the proactive safety assurance capability of switchgear operation.

[0023] 2. This invention automatically collects negative sample cases through a false alarm feedback mechanism, uses a difference analysis engine to compare data characteristics under normal and false alarm states, and dynamically adjusts the weight allocation of various inducing parameters and warning thresholds. This self-correcting mechanism enables the system to adapt to changes in different operating environments and equipment states, effectively reducing false alarms and missed alarms. At the same time, by continuously accumulating a fault case library to optimize diagnostic rules, the accuracy and reliability of the system's judgment continuously improve with the extension of operating time, providing continuously evolving intelligent protection for the long-term stable operation of the switchgear. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0025] Figure 1 This is a flowchart illustrating a method and system for monitoring fire and explosion-proof anomalies in an intelligent switchgear.

[0026] Figure 2 This is a schematic diagram of the system architecture of an intelligent switchgear anti-fire and explosion anomaly monitoring method and system. Detailed Implementation

[0027] To more clearly illustrate the overall concept of the present invention, a detailed description will be provided below with reference to the accompanying drawings and examples.

[0028] In the description of this invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and 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 this invention.

[0029] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0030] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0031] In this invention, unless otherwise expressly specified and limited, the first feature "on" or "below" the second feature may be in direct contact with the first and second features, or indirect contact through an intermediate medium. In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0032] like Figures 1 to 2 As shown, this invention proposes an intelligent method for monitoring fire and explosion anomalies in switchgear, comprising the following steps:

[0033] Step 1: Multi-source sensing and data aggregation. This involves comprehensive condition monitoring using multi-source sensors deployed at key nodes within the switchgear. The core of data acquisition is covering the main types of faults leading to combustion and explosion. This includes real-time monitoring of overheating risks using wireless passive temperature sensors installed at high-voltage points such as circuit breaker contacts; capturing electromagnetic and acoustic signals of partial discharge using UHF and ultrasonic sensors to assess insulation degradation; monitoring the concentration of characteristic gases from the decomposition of insulating materials, such as hydrogen and carbon monoxide, using electrochemical gas sensors; and using high-speed arc flash sensors as a last line of defense against internal arcing faults. All this data is aggregated through an edge computing gateway within the cabinet. This gateway performs initial data cleaning, time synchronization, and caching before uploading it to the central analysis system via the industrial network, providing a unified, high-quality data foundation for subsequent analysis.

[0034] The second step is data association modeling and archiving. This involves establishing the relationship between result and cause data, defining parameters that directly characterize the fault results, such as temperature, characteristic gas concentration, and partial discharge intensity, as result data, and defining potential cause parameters, such as current, voltage, and humidity, as cause data. The result data is then grouped and aggregated to collect data from the cause data. For example, a temperature database is set up to collect current, voltage, and humidity corresponding to different temperatures.

[0035] Step 3: Dynamic correlation analysis and weight allocation. This step calculates the importance of different inducing factors under the same result data. Within the same result data, different inducing factors corresponding to different state result data are selected, and a circuit diagram is generated for curvature calculation. For example, a temperature change curve is generated with time as the horizontal axis and temperature as the vertical axis, and a voltage change curve is generated with time as the horizontal axis and voltage as the vertical axis. The curvature of the temperature change curve and the curvature of the voltage change curve are calculated respectively. In this way, the curvature of the change curves of other inducing factors is calculated and compared with the curvature of the temperature change curve. The closer it is to the curvature, the higher its influence on temperature change. Thus, the weight of each inducing factor data in the same result data is determined. The core of this step is to quantify the dynamic correlation strength between the inducing factor and the result by analyzing the similarity of the shape of the data curve (not just the numerical value), which can more sensitively capture those key inducing factors that evolve synchronously with the fault.

[0036] Step 4: Time series modeling and prediction weight calculation. For the causal data corresponding to the same result data, since some results have a time delay, such as the increase in voltage in the first one or two minutes leading to the increase in temperature in the next one or two minutes, the time-sensitivity of the causal data is extended. For example, the causal data will cause the subsequent result data to increase or decrease.

[0037] The system will introduce time series analysis algorithms to automatically find the optimal time lag interval between different inducing data and result data, and assign prediction weights to these lagged inducing data, so that the system can predict the trend of result data changes in the future based on the changes of the current inducing factors.

[0038] Step 5: Real-time Risk Prediction and Early Warning. The detected result data or triggering data is transmitted to the weighted model established in Step 3 and the timeliness model established in Step 4 for real-time risk assessment. The system no longer passively waits for result data to exceed limits; instead, it continuously calculates a "risk index" that integrates changes in all high-weight, timeliness-sensitive triggering data. When this index indicates that the system state is moving towards a known failure mode, the system will trigger an early warning even if all result data remains within the normal range.

[0039] Step 6: Misjudgment feedback and model self-correction. When the system detects significant fluctuations in one or more triggering data and issues an early warning accordingly, if no expected changes in the corresponding result data are observed within the set extended time period, the system will determine that the early warning is a misjudgment or an oversensitive warning.

[0040] When the above situation occurs multiple times, the specific number of times can be set to a threshold, such as two times. After two false positives, the system will automatically record the complete data sequence of this event to form a negative sample case.

[0041] Based on negative sample cases, the system will activate a self-correction mechanism. First, it will lower the weight of the specific triggering data that caused the misjudgment, so that the weight of the result data will not easily trigger a high-level alarm in similar scenarios in the future. Then, it will continue to compare the second weight when the result data appears, increase the weight value of the second weight, and make another judgment. If the second weight still results in a misjudgment, it is necessary to comprehensively consider all triggering data, record all data in the negative sample case, and compare them with the data when the failure occurred in the database before, find the cases where the misjudged data is the same, and find the triggering data with different values ​​under the same misjudged data. Combine the triggering data with the previous misjudged data for monitoring to complete the judgment update.

[0042] The system architecture based on the above-mentioned intelligent switchgear fire and explosion anomaly monitoring method includes a data acquisition module, a data storage and management module, a correlation analysis module, a time series prediction module, a risk assessment and early warning module, a communication interface module, and a human-computer interaction module.

[0043] The data acquisition module consists of multiple types of sensor units, a signal conditioning unit, and an edge computing node. The sensor units include a temperature sensor array, a characteristic gas sensor group, a partial discharge detector, a current sensor, a voltage transformer, and a humidity sensor. The signal conditioning unit completes analog-to-digital conversion, filtering and noise reduction, and standardized output. The edge computing node is deployed in the switch cabinet field to realize raw data compression and preprocessing, and transmits it to the upper layer through industrial Ethernet or RS485 bus.

[0044] The data storage and management module constructs a hybrid storage architecture of time-series database and relational database. The time-series database is responsible for storing real-time monitoring data streams, and tables are created according to result data and cause data. The result data table uses temperature, gas concentration, and partial discharge intensity as primary key indexes, while the cause data table uses current, voltage, and humidity as fields, and establishes relationships through timestamps. The relational database is used to store weight model parameters, early warning records, negative sample case library, and system configuration information. This module has a built-in data archiving engine that performs data grouping and set operations according to preset periods to build a cause data mapping table with result data as the core.

[0045] The association analysis module implements dynamic association weight calculation. It uses a dynamic time warping algorithm to calculate the curve similarity between the cause data sequence and the result data sequence, and outputs the association weight value in the 0-1 interval. It has a weight update unit that automatically triggers the weight recalculation process when a new fault case is detected. It uses the gradient descent method to optimize the weight allocation and configures a threshold manager to filter weak association causes below the sensitivity threshold, retaining only the set of high-weight causes for subsequent prediction.

[0046] The time series prediction module integrates an autoregressive integral moving average model and a long short-term memory neural network algorithm. The lag analysis engine automatically performs cross-correlation function calculations to determine the optimal time lag interval of each inducing data relative to the result data and generates a lag parameter table. The weighted time extension unit performs time shifting on historical inducing data according to the lag interval to construct a prediction feature vector. The prediction model trainer adopts an online learning method, continuously incorporating newly collected data into the training set, updating the model parameters every 24 hours, and outputting the predicted value and confidence interval of the result data for the next 60 seconds.

[0047] The risk assessment and early warning module includes a risk index calculator and a graded early warning generator. The risk index calculator reads the current trigger data in real time, substitutes it into the weight model and the timeliness model, and calculates the comprehensive risk score. The graded early warning generator sets three threshold levels (attention, abnormal, and emergency). When the threshold is exceeded, it triggers the sound and light alarm and the SMS gateway, and sends a trip or ventilation command to the execution layer at the same time.

[0048] The self-learning optimization module consists of a misjudgment identification unit and a negative sample processing unit. After monitoring and issuing warnings, the misjudgment identification unit monitors and issues warnings for changes in the result data within a set time period. If the expected fluctuation range is not reached, it is judged as a misjudgment, and the complete data sequence is automatically recorded and marked as a negative sample. When the negative sample processing unit accumulates negative samples to the number of times threshold (default 2 times), it starts the difference analysis engine, extracts the differences in the causal data features between misjudged cases and positive samples, generates avoidance rules, and injects them into the correlation analysis module to achieve model self-correction.

[0049] The communication interface module enables connection with the substation's integrated automation system, reserves an MQTT interface for cloud data synchronization, and employs a hardware watchdog and dual redundant communication links to ensure reliability.

[0050] The human-computer interaction module is based on a web architecture and provides real-time monitoring screens, trend curve displays, early warning information lists, model parameter configuration, and historical data query functions, and supports mobile access.

[0051] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0052] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for monitoring fire and explosion anomalies in intelligent switchgear, characterized in that: Includes the following steps: Step 1: Multi-source sensing and data aggregation, using multi-source sensors deployed at key nodes inside the switch cabinet for comprehensive status detection; Step 2: Data association modeling and archiving steps, establish the association between result data and cause data, define parameters that directly characterize the failure results as result data, define potential cause parameters as cause data, and construct a cause data mapping table with result data as the core. Step 3: Dynamic correlation analysis and weight allocation. By calculating the curve similarity between each causal data sequence and the result data sequence, the contribution weight of each causal parameter to the fault result is quantified. The importance of different causal data under the same result data is calculated. In the same result data, different causal data corresponding to different state result data are selected, and a circuit diagram is generated for curvature calculation. The dynamic correlation strength between causal and result is quantified by analyzing the morphological similarity of the data curves. Step 4: Time series modeling and prediction weight calculation. The time series analysis method is used to determine the optimal time lag interval between each inducing factor data and the result data. The time series analysis algorithm automatically finds the optimal time lag interval between different inducing factor data and the result data, and assigns prediction weights to these lagged inducing factor data to establish a prediction model. Step 5: Real-time risk prediction and early warning. Based on the weighting model and the timeliness model, a comprehensive risk index is calculated to achieve tiered early warning. Step 6: Misjudgment Feedback and Model Self-Correction. After detecting significant fluctuations in one or more triggering data and issuing an early warning, if no expected change in the corresponding result data is observed within the set extended timeframe, the system will determine this warning as a misjudgment, resulting in an incorrect judgment sample. This sample is collected as a negative sample case, and the weight allocation and warning threshold are dynamically adjusted. If the number of misjudgments exceeds the set threshold, it is judged as a negative sample case. In the model self-correction process, the weight of the specific triggering data that caused the misjudgment is reduced, and the weight of the second weight is increased. The case is judged again. If the second weight still results in a misjudgment, the data records in the negative sample case are compared with the data in the database when the fault occurred. The system identifies cases where the misjudged data is the same and finds the triggering data with different values ​​under the same misjudged data. This triggering data is then combined with the previous misjudged data for monitoring, completing the judgment update and realizing the model's self-learning optimization.

2. An intelligent switchgear fire and explosion prevention anomaly monitoring system, which is applied to the intelligent switchgear fire and explosion prevention anomaly monitoring method described in claim 1, characterized in that: It includes a data acquisition module, a data storage and management module, a correlation analysis module, a time series prediction module, a risk assessment and early warning module, a communication interface module, and a human-computer interaction module.

3. The intelligent switchgear fire and explosion prevention anomaly monitoring system according to claim 2, characterized in that: The data acquisition module consists of multiple types of sensor units, signal conditioning units, and edge computing nodes, while the data storage and management module constructs a hybrid storage architecture of time-series databases and relational databases.

4. The intelligent switchgear fire and explosion prevention anomaly monitoring system according to claim 2, characterized in that: The association analysis module implements the function of dynamic association weight calculation. It uses the dynamic time warping algorithm to calculate the curve similarity between the causal data sequence and the result data sequence, outputs the association weight value in the 0-1 interval, and has a weight update unit.

5. The intelligent switchgear fire and explosion prevention anomaly monitoring system according to claim 2, characterized in that: The time series prediction module integrates an autoregressive integral moving average model and a long short-term memory neural network algorithm. The lag analysis engine automatically performs cross-correlation function calculations to determine the optimal time lag interval of each causal data relative to the result data and generates a lag parameter table. The weighted time extension unit performs time shifting on the historical causal data according to the lag interval and constructs a prediction feature vector.

6. The intelligent switchgear fire and explosion prevention anomaly monitoring system according to claim 2, characterized in that: The risk assessment and early warning module includes a risk index calculator and a graded early warning generator. The self-learning optimization module consists of a misjudgment identification unit and a negative sample processing unit. The communication interface module enables connection with the substation integrated automation system. The human-machine interaction module is based on a Web architecture and provides real-time monitoring screens, trend curve displays, early warning information lists, model parameter configuration, and historical data query functions, supporting mobile access.

Images