Intelligent fire extinguishing system for transformer

Abstract

According to the intelligent fire extinguishing system for the transformer, internal and external data of the transformer are collected in real time through a sensor module, the collected data are preprocessed through a data processing and model module, and a large model architecture based on Transform and a multi-head self-attention mechanism are adopted; the current temperature state and the future temperature change trend of the transformer are predicted in real time, risk assessment and anomaly detection are carried out, multi-stage temperature threshold values are set in the system, real-time data processing and model optimization are achieved through an edge computing and cloud computing cooperation module, the response speed and prediction accuracy of the system are improved, and the system reliability is improved. The infrared monitoring module monitors the temperature change and the fire condition outside the transformer in real time and accurately detects a fire or an abnormal high-temperature area, and when the fire is detected or a dangerous-level temperature early warning is received, the fire extinguishing module starts the non-pressure-storage fire extinguisher, sprays a fire extinguishing agent and rapidly extinguishes a fire, so that the safety and reliability of the transformer are enhanced, and the safety of the transformer is improved. And the operation cost is reduced, and the application prospect is wide.

Description

Technical Field

[0001] The invention relates to an intelligent fire extinguishing system for a transformer, and in particular to an intelligent fire extinguishing system for a transformer applied in the technical field of transformer fire extinguishing. Background Art

[0002] With the development of the power system and the continuous increase in grid load, the safe operation of transformers, as key equipment for power transmission and distribution, is of vital importance. During long-term operation, transformers may overheat or even cause fires due to factors such as overload, insulation aging, and increased ambient temperature, causing significant economic losses and safety hazards.

[0003] Existing transformer fire extinguishing technology mainly relies on passive fire detection and fire extinguishing measures. Temperature sensors are usually used to detect the temperature of the transformer. Once the temperature exceeds the set threshold, the fire extinguishing device is triggered. However, this method has the following problems: single monitoring means and limited data acquisition: Traditional temperature monitoring mainly relies on a single temperature sensor, which cannot fully obtain the internal and external operating status and environmental information of the transformer, resulting in incomplete monitoring data and lack of intelligent prediction capabilities: Existing systems usually take measures only after the temperature exceeds the threshold, lack of prediction of temperature trends, and cannot provide early warning, which can easily lead to untimely fire extinguishing, slow response speed, and low fire extinguishing efficiency: Due to the lack of real-time risk assessment and rapid response mechanism, the activation of the fire extinguishing device may be delayed, increasing the risk of fire spread, and the reliability of the fire extinguishing system is insufficient: The fire extinguishing device lacks self-inspection and fault handling mechanisms, and may fail at critical moments, and the fire extinguishing effect cannot be guaranteed.

[0004] In response to the above problems, some technical solutions have been proposed in the industry to try to improve them.

[0005] Chinese invention patent CN112103991B discloses a multi-supply area flexible interconnection system, including a regional energy dispatching station, a user-side intelligent interconnection collection terminal, an on-grid and off-grid intelligent control device, and a mobile energy storage supply station. This solution can quickly adjust the ambient temperature of the energy storage system and automatically extinguish fires, ensuring the safe, stable, and efficient operation of the energy storage system.

[0006] Chinese invention patent CN112072537B discloses an active early warning system for overvoltage in a distribution network based on ubiquitous Internet of Things technology, including a distribution network detection system and a distribution cabinet, which can realize real-time detection of equipment and quickly take fire extinguishing and cooling measures when a fault occurs.

[0007] The above design has improved the safety and response speed of the system to a certain extent by introducing intelligent monitoring and automatic fire extinguishing technology, but there are still certain limitations. The above patents are mainly aimed at the early warning and fire extinguishing of overvoltage in energy storage systems and distribution networks. The research on the dedicated fire extinguishing system for transformers is not in-depth enough, and there is a lack of deep integration and analysis of multi-source data. It fails to fully utilize advanced artificial intelligence models for temperature prediction and risk assessment, and has not established a multi-level early warning mechanism. It is impossible to take corresponding response measures according to different risk levels. The startup and control process of the fire extinguishing device is relatively simple, and there is a lack of self-inspection and fault handling mechanism. The selection and injection process of the fire extinguishing agent are uncontrollable, which may lead to incomplete fire extinguishing or secondary damage to the equipment. Summary of the invention

[0008] In view of the above-mentioned prior art, the technical problem to be solved by the present invention is to provide an intelligent fire extinguishing system for a transformer, which can realize comprehensive real-time monitoring of the operating status of the transformer, use artificial intelligence models to perform high-precision temperature prediction and risk assessment, establish a multi-level early warning mechanism, issue early warnings in time and take corresponding response measures, quickly and effectively extinguish fires, and at the same time have self-checking and fault handling functions to ensure the reliability and safety of the system.

[0009] To solve the above problems, the present invention provides an intelligent fire extinguishing system for transformers, including a sensor module, a data processing and model module, an edge computing and cloud computing collaboration module, a prediction and early warning module, an infrared monitoring module, a fire extinguishing module, and a self-checking and fault handling module;

[0010] A sensor module is used to collect internal and external data of the transformer, including but not limited to power change, input power and output power difference, transformer internal temperature, voltage, current, ambient temperature, humidity and air pressure;

[0011] The data processing and model module is connected to the sensor module to pre-process the collected data, including data cleaning, normalization and feature extraction. Based on the Transformer large model architecture and self-attention mechanism, the data is converted into a vector representation that can be understood by the model. The multi-head self-attention mechanism is used to analyze the various factors that affect the transformer temperature, and the current temperature state of the transformer and future temperature change trends are predicted in real time.

[0012] The edge computing and cloud computing collaborative module connects the data processing and model modules. The edge computing module is used for real-time processing and response, and performs temperature prediction and early warning. The edge computing and cloud computing collaborative module includes an edge computing module and a cloud computing module. The cloud computing module is used for long-term data storage, backup, and model training optimization. The edge computing module and the cloud computing module work together through an efficient data communication protocol.

[0013] The prediction and warning module connects the data processing and model module and the edge computing and cloud computing collaboration module. According to the temperature prediction results, when the predicted temperature may exceed the set threshold, it issues a multi-level warning signal and activates the fire extinguishing module to enter the preparatory working state;

[0014] Infrared monitoring module, including infrared camera and visible light camera, monitors temperature changes and fire conditions outside the transformer in real time, and detects whether a fire or abnormally high temperature area occurs by combining thermal imaging and visible light images;

[0015] The fire extinguishing module includes a non-pressure storage type fire extinguisher, which is connected to the infrared monitoring module and the prediction and warning module. When the infrared monitoring module detects a fire or receives a temperature warning of a dangerous level, the fire extinguisher is activated to extinguish the fire. The fire extinguisher generates high-pressure gas through an internal gas generator and sprays a fire extinguishing agent. The fire extinguishing agent is suitable for electrical fires and does not damage the transformer;

[0016] The self-check and fault handling module is connected to the fire extinguishing module. When an early warning signal is issued, the self-check mechanism of the fire extinguisher is triggered to detect the gas generator status, fire extinguishing agent capacity, nozzle patency and electrical connection integrity of the fire extinguisher to ensure that the fire extinguisher is in normal working condition. If the fire extinguisher is detected to be invalid or abnormal, an alarm signal will be issued in time to notify the maintenance personnel.

[0017] In the above-mentioned intelligent fire extinguishing system for transformers, a variety of sensors, edge computing and cloud computing collaboration, intelligent analysis and prediction based on Transformer model, and an intelligent fire extinguishing system with multi-level early warning and automatic fire extinguishing functions are integrated. The system can collect multi-source data inside and outside the transformer in real time, integrate external information such as weather and power grid load, and use a deep learning model with a multi-head self-attention mechanism to accurately predict the current temperature state and future temperature trend of the transformer, and identify potential overheating risks in advance. At the same time, the system sets a multi-level temperature threshold, and sends out corresponding early warning signals according to different risk levels, and takes corresponding measures, such as strengthening monitoring, preparing for fire extinguishing, or immediately starting the fire extinguishing module. The fire extinguishing module adopts a non-pressure storage fire extinguisher, and the fire extinguishing agent can be selected according to actual needs. The injection pressure and time are adjustable to avoid secondary damage to the equipment. The self-checking and fault handling module can be automatically triggered when the early warning is issued, and the key components of the fire extinguisher are detected to ensure the reliability of the fire extinguishing system. Through the above-mentioned technical scheme, the present invention effectively overcomes the shortcomings of the prior art, realizes intelligent monitoring, prediction and fire extinguishing processing of transformers, improves the safety and reliability of transformer operation, and has important practical value and broad application prospects.

[0018] As a further improvement of the present application, the sensor module includes a power sensor, which is used to measure the input power and output power of the transformer and calculate the power difference to evaluate the heat generation of the transformer; temperature sensors are distributed in multiple key parts of the transformer, including windings, oil tanks and radiators, to obtain comprehensive temperature information.

[0019] As a further improvement of this application, the data processing and model module adopts a Transformer-based large model architecture and uses a multi-head self-attention mechanism to analyze the factors affecting the transformer temperature from multiple angles such as load changes, ambient temperature, heat dissipation capacity, weather conditions, and grid load data, thereby improving the accuracy and reliability of temperature prediction.

[0020] As a further improvement of the present application, the edge computing module in the edge computing and cloud computing collaborative module includes a high-performance processor and storage unit, which can execute complex prediction and early warning algorithms locally in real time; the cloud computing module is based on a cloud server, using big data analysis and machine learning technology to conduct in-depth mining of historical data and continuously optimize the prediction model.

[0021] As another improvement of the present application, the prediction and early warning module sets multiple levels of temperature thresholds, including normal level, attention level, warning level and danger level. When the predicted temperature exceeds different thresholds, early warning signals of corresponding levels are issued respectively. The normal level maintains monitoring, the attention level strengthens monitoring, the warning level prepares for fire extinguishing, and the danger level immediately activates the fire extinguishing module.

[0022] As another improved supplement to the present application, the fire extinguishing agent of the non-pressure storage fire extinguisher in the fire extinguishing module is an inert gas, clean gas or dry powder suitable for electrical fire extinguishing, including but not limited to heptafluoropropane, sulfur hexafluoride or carbon dioxide. The starting process of the fire extinguisher is controlled, and the injection pressure and time are adjustable to avoid secondary damage to the transformer and surrounding equipment.

[0023] As another improved supplement to the present application, the self-test and fault handling module can trigger the self-test of the fire extinguisher on a regular, manual and early warning basis, and the detection contents include but are not limited to: the integrity of the power supply and starting circuit of the gas generator; the remaining amount and effectiveness of the fire extinguishing agent; the patency and sealing of the nozzle and pipeline; the communication and control signal transmission status of the system; if any abnormality is detected, an alarm signal is immediately issued, and the fault information is sent to the remote monitoring center or the maintenance personnel's mobile device through the communication module.

[0024] As another improvement of the present application, the external data integration module obtains real-time weather data, including temperature, humidity, wind speed, rainfall and air pressure, as well as the load, power supply frequency and voltage fluctuation of the local power grid, and provides these data together with the data of the sensor module to the data processing and model module to optimize the temperature prediction model; the system has online learning capabilities and can continuously adjust and optimize model parameters based on newly acquired data and operating conditions to improve the accuracy of predictions and warnings.

[0025] An intelligent fire extinguishing system for transformers, characterized in that: its working method comprises the following steps:

[0026] S1: Data collection, collecting internal and external data of the transformer through the sensor module, including power change, input power and output power difference, transformer internal temperature, voltage, current, ambient temperature, humidity and air pressure;

[0027] S2: Data preprocessing: In the data processing and model module, the collected data is cleaned, normalized, and features are extracted to convert the data into a vector representation that can be understood by the model;

[0028] S3: Temperature prediction

[0029] S3.1: Model input preparation

[0030] Input the vectorized data preprocessed in step S2 into the data processing and model module, the data including but not limited to: power change data of the transformer; difference between input power and output power; temperature data of internal temperature, including temperature of winding, oil tank and radiator; voltage and current data; environmental parameters of ambient temperature, humidity and air pressure; weather condition data, such as wind speed, rainfall and atmospheric pressure; local power grid load data and voltage fluctuation;

[0031] S3.2: Feature association analysis

[0032] S3.2.1: Use the self-attention mechanism to perform feature correlation analysis on the input data;

[0033] S3.2.2: Calculate the correlation weights between the features and identify the key factors that have a significant impact on the transformer temperature, such as load fluctuations and ambient temperature changes.

[0034] S3.3: Multi-head self-attention mechanism processing

[0035] S3.3.1: In the Transformer large model architecture, a multi-head self-attention mechanism is used to process features from different subspaces in parallel;

[0036] S3.3.2: Each attention head focuses on different feature combinations to capture complex nonlinear relationships and interactions, such as the correlation between power difference and internal temperature, and the impact of ambient humidity on heat dissipation efficiency.

[0037] S3.4: Deep feature extraction and representation

[0038] S3.4.1: Deeply extract features processed by the multi-head self-attention mechanism by stacking multiple layers of encoders;

[0039] S3.4.2: Generate a high-dimensional representation of the transformer temperature state that captures both the time series and spatial distribution characteristics.

[0040] S3.5: Current temperature status prediction

[0041] S3.5.1: Input the high-dimensional feature representation into the fully connected layer or regression layer to calculate the current temperature values ​​of each key part of the transformer;

[0042] S3.5.2: Temperature distribution of the output transformer as a whole, identifying possible overheating areas;

[0043] S3.6: Prediction of future temperature trends

[0044] S3.6.1: Based on the historical data and time series patterns learned by the model, predict the temperature change trend within a specified time range in the future;

[0045] S3.6.2: Generate future temperature change curves to predict possible temperature peaks and rates of change;

[0046] S3.7: Risk Assessment and Anomaly Detection

[0047] S3.7.1: Perform a risk assessment on the predicted temperature results and calculate the probability and likelihood of temperature exceeding the limit;

[0048] S3.7.2: Use set temperature thresholds to perform anomaly detection and identify potential overheating risks and abnormal temperature rise conditions;

[0049] S3.8: Prediction result output

[0050] S3.8.1: Integrate current temperature status, future temperature trends, risk assessment, and anomaly detection results;

[0051] S3.8.2: Output the integrated forecast results to the forecast and warning module to provide a basis for subsequent warning and processing;

[0052] S3.8.3: At the same time, key prediction data is transmitted to the edge computing and cloud computing collaborative module for data backup and model optimization.

[0053] S4: Edge and cloud computing work together to perform temperature prediction and early warning in real time through edge computing modules and cloud computing modules, and use cloud computing modules to perform long-term data storage, backup, and model training optimization;

[0054] S5: Early warning mechanism: According to the temperature prediction results, when the predicted temperature may exceed the set multi-level threshold, the prediction and early warning module sends out an early warning signal of the corresponding level and activates the fire extinguishing module to enter the preparatory working state;

[0055] S6: Infrared monitoring: The infrared camera and visible light camera of the infrared monitoring module monitor the temperature changes and fire conditions outside the transformer in real time to detect whether there is a fire or abnormally high temperature area;

[0056] S7: Self-check and fault handling. When the warning signal is issued, the self-check and fault handling module triggers the self-check mechanism of the fire extinguisher to detect the gas generator status, fire extinguishing agent capacity, nozzle patency and electrical connection integrity. If the fire extinguisher is detected to be invalid or abnormal, an alarm signal is issued in time and the maintenance personnel are notified to handle it;

[0057] S8: Fire extinguishing process. When the infrared monitoring module detects a fire or receives a temperature warning of a dangerous level, the fire extinguishing module activates the non-pressure storage fire extinguisher. The internal gas generator produces high-pressure gas and sprays a fire extinguishing agent suitable for electrical fires to extinguish the transformer.

[0058] In summary, this application has the following beneficial effects:

[0059] 1. Realize real-time monitoring of transformer operation status and high-precision temperature prediction. Through the sensor module, collect key parameters such as transformer power change, input power and output power difference, internal temperature, voltage, current, etc. in real time, obtain ambient temperature, humidity, air pressure, weather data (such as wind speed, rainfall) and grid load data, integrate internal and external data, enrich the input of the model, and improve the accuracy of temperature prediction. The Transformer model in deep learning has the ability to process long sequence data and capture global dependencies. The multi-head self-attention mechanism processes features in parallel from different subspaces, captures complex nonlinear relationships and feature interactions, and improves the model's sensitivity to temperature changes. The model can predict the current temperature state and future temperature change trend of the transformer in real time, identify possible overheating risks in advance, and calculate the probability of temperature exceeding the limit through risk assessment and anomaly detection, providing a reliable early warning basis.

[0060] 2. Build a multi-level early warning mechanism to improve risk management capabilities. Set multi-level temperature thresholds. When the temperature is within a safe range, the system monitors normally. When the temperature approaches the threshold, the system sends a warning signal and strengthens monitoring. When the temperature exceeds the warning threshold, the system sends a warning signal and prepares to extinguish the fire. When the temperature far exceeds the threshold, the system sends a danger warning signal and immediately starts the fire extinguishing module. According to different risk levels, the system takes corresponding measures to avoid overreaction or underreaction, and takes preventive measures before the risk escalates to prevent accidents. The warning information is promptly conveyed to the operation and maintenance personnel and the management center through sound and light alarms, text messages, emails and mobile applications. The warning information contains temperature data, risk levels, recommended measures, etc., which is convenient for decision-making.

[0061] 3. Improve the reliability and effectiveness of the fire extinguishing system. The internal gas generator produces high-pressure gas. The fire extinguisher can be started in a short time to quickly control the fire, avoiding the pressure leakage problem that may occur in the pressure storage fire extinguisher, and improving the reliability of the system. Inert gas, clean gas or dry powder fire extinguishing agent, such as heptafluoropropane, sulfur hexafluoride, carbon dioxide, etc., can be selected according to actual needs. The fire extinguishing agent has no damage to transformers and electrical equipment, avoiding secondary damage. The injection pressure and time are adjustable to meet the needs of different fire conditions, accurately control the amount of fire extinguishing agent, and prevent resource waste and equipment damage.

[0062] 4. Enhance the self-inspection and fault handling capabilities to ensure stable operation of the system. Self-inspection can be triggered regularly, manually and by early warning to ensure that the fire extinguisher is always in normal condition, including gas generator, power supply, starting circuit, fire extinguishing agent capacity, nozzle patency, electrical connection, etc. Comprehensively check the key components of the system. When an abnormality is detected, the system immediately sends an alarm signal to prompt on-site personnel. Through the communication module, the fault information is sent to the remote monitoring center or maintenance personnel for rapid response. Through the self-inspection mechanism, potential faults can be discovered in advance to avoid equipment failure at critical moments, handle faults in a timely manner, and reduce system downtime and maintenance time.

[0063] 5. Edge computing and cloud computing work together to improve system performance and intelligence, perform real-time processing and response locally, perform temperature prediction and early warning, meet low latency requirements, reduce data transmission volume, and improve overall system efficiency. The cloud computing module stores historical data for a long time to prevent data loss, and uses the powerful computing power of the cloud to regularly train and optimize models to improve prediction performance. The edge and cloud work together through efficient data communication protocols to ensure the consistency of data and models. According to the network and computing resource conditions, tasks are intelligently scheduled to improve system resource utilization.

[0064] 6. It has online learning capabilities to improve the model's adaptability and prediction accuracy. The model continuously adjusts and optimizes parameters based on newly acquired data and operating conditions to adapt to environmental changes. It can handle new fault modes and abnormal conditions and improve prediction accuracy. The external data integration module obtains weather data, power grid load data, etc., and integrates them with sensor data to improve the generalization ability of the model. The introduction of multi-source data enables the model to consider influencing factors more comprehensively and make prediction results more reliable. Through continuous optimization of the model, false alarms are reduced and the accuracy of early warnings is improved. Accurate predictions and early warnings reduce the possibility of missed alarms and ensure safe operation of transformers.

[0065] 7. The system is highly intelligent and automated. From data collection to early warning, no human intervention is required, reducing human errors. In dangerous situations, the system automatically activates the fire extinguisher to quickly control the fire. The system makes intelligent decisions based on real-time data and model analysis, provides detailed early warning and fault information, and assists operation and maintenance personnel in formulating maintenance strategies. Automated monitoring and maintenance reduces dependence on manual inspections. The system monitors and processes in real time, improving fault response speed.

[0066] 8. Ensure the safe operation of transformers and power systems. Through real-time monitoring and accurate prediction, potential risks can be discovered in advance to prevent them. When a fire occurs, the system responds quickly to reduce accident losses, promptly handles overheating and fire problems, and avoids power outages caused by transformer failures. The safe operation of transformers helps maintain the stability of the power system, prevents transformers from being damaged by overheating or fire, extends equipment life, and reduces maintenance costs through intelligent monitoring and maintenance.

[0067] 9. Reduce operating costs, improve economic benefits, reduce the need for on-site personnel, reduce labor costs, reduce the frequency of on-site inspections through remote monitoring and management, arrange maintenance plans reasonably according to system warnings and self-inspection results, avoid unnecessary maintenance, reduce failures and downtime, improve the use efficiency of transformers, and reduce energy loss by optimizing the operating status of transformers. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 For the overall interaction of this application Figure 1 ;

[0069] Figure 2 For the overall interaction of this application Figure 2 ;

[0070] Figure 3 For the overall interaction of this application Figure 3 ;

[0071] Figure 4 This is the feature association analysis formula diagram for this application. DETAILED DESCRIPTION

[0072] Three embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0073] Example 1

[0074] like Figures 1 to 3 As shown, this embodiment provides an intelligent fire extinguishing system for a transformer, which is intended to monitor the operating status of the transformer in real time, predict potential overheating risks, provide timely warnings and automatically start the fire extinguishing device when necessary, so as to improve the safety and reliability of the transformer operation.

[0075] The intelligent fire extinguishing system includes the following main modules:

[0076] Sensor Module:

[0077] Power sensor: installed at the input and output ends of the transformer to measure input power and output power, calculate the power difference to evaluate the heat generated by the transformer, temperature sensor: distributed in key parts of the transformer, including windings, oil tanks and radiators, to obtain comprehensive temperature information, voltage and current sensors: monitor the voltage and current parameters of the transformer, environmental sensors: installed around the transformer to collect external data such as ambient temperature, humidity and air pressure.

[0078] Data processing and model modules:

[0079] Data preprocessing unit: cleans, normalizes and extracts features from the data collected by the sensor. Model calculation unit: based on the Transformer large model architecture, it uses the multi-head self-attention mechanism to analyze the factors affecting the transformer temperature from multiple angles such as load changes, ambient temperature, and heat dissipation capacity, and predicts the current temperature state of the transformer and future temperature change trends in real time.

[0080] Edge computing and cloud computing collaboration module:

[0081] Edge computing module: includes high-performance processors and storage units, deployed at the transformer site, responsible for real-time processing of sensor data, temperature prediction and early warning. Cloud computing module: based on cloud servers, responsible for long-term data storage, backup and model training optimization, and works with the edge computing module through efficient data communication protocols.

[0082] Forecast and warning module:

[0083] According to the temperature prediction results, the temperature threshold is set. When the predicted temperature may exceed the set threshold, a multi-level warning signal is issued. The warning signal is transmitted to the operation and maintenance personnel through the sound and light alarm or the communication network, and the fire extinguishing module is activated to enter the preparatory working state.

[0084] Infrared monitoring module:

[0085] Infrared camera: Real-time monitoring of thermal imaging of the transformer exterior to detect abnormally high temperature areas.

[0086] Visible light camera: provides visible light images of the outside of the transformer to assist in judging the fire situation. The combination of the two enables real-time monitoring of the transformer's external temperature changes and fire conditions.

[0087] Fire extinguishing module:

[0088] Non-pressure storage fire extinguisher: High-pressure gas is generated through the internal gas generator to spray fire extinguishing agents suitable for electrical fires (such as heptafluoropropane). The fire extinguisher is connected to the infrared monitoring module and the prediction and warning module. When a fire is detected or a temperature warning of a dangerous level is received, the fire extinguisher is automatically activated to extinguish the fire.

[0089] Self-check and troubleshooting module:

[0090] When a warning signal is issued, the fire extinguisher's self-check mechanism is triggered to detect the gas generator status, fire extinguishing agent capacity, nozzle patency and electrical connection integrity. If the fire extinguisher is detected to be ineffective or abnormal, an alarm signal is issued in time and maintenance personnel are notified to handle it.

[0091] Working process

[0092] The sensor module continuously collects internal and external data of the transformer, including power, temperature, voltage, current and environmental parameters, and transmits the collected data to the data preprocessing unit of the data processing and model module.

[0093] The data is cleaned to remove outliers and noise; normalized to unify the data scale; feature extraction is performed to extract key influencing factors. The preprocessed data is input into the model calculation unit, and the Transformer-based large model architecture and multi-head self-attention mechanism are used to analyze the impact of each feature on temperature.

[0094] Real-time prediction of the current temperature status and future temperature change trend of the transformer. The prediction and early warning module compares the prediction results with the set temperature threshold: the temperature is within the safe range and the system is monitoring normally. Attention level: the temperature is close to the threshold, a warning signal is issued, and monitoring is strengthened. Warning level: the temperature exceeds the warning threshold, a warning signal is issued, and fire extinguishing is prepared. Danger level: the temperature far exceeds the threshold, a danger warning signal is issued, and the fire extinguishing module is immediately started.

[0095] The edge computing module processes data in real time on site, performs temperature prediction and early warning, and ensures low-latency response. Key data and historical data are transmitted to the cloud computing module through a secure communication protocol. The cloud computing module stores and backs up data for a long time and uses machine learning technology to train and optimize the model to improve prediction accuracy.

[0096] The infrared monitoring module continuously monitors the external temperature and fire conditions of the transformer, and detects anomalies in real time by combining thermal imaging and visible light images. When a fire is detected or a warning signal of a dangerous level is received, the non-pressure storage fire extinguisher of the fire extinguishing module is activated: the gas generator generates high-pressure gas and quickly sprays the fire extinguishing agent onto the surface of the transformer and the internal overheating area. The fire extinguishing agent is a gas that does not harm the transformer, and quickly suppresses the fire and prevents the fire from spreading.

[0097] When the early warning signal is issued, the self-test and fault handling module starts the self-test mechanism of the fire extinguisher: detects the power supply and starting circuit integrity of the gas generator, checks the remaining amount and effectiveness of the fire extinguishing agent, detects the patency and sealing of the nozzle and pipeline, and checks the communication and control signal transmission status of the system.

[0098] If any abnormality is detected, an alarm signal will be immediately issued and maintenance personnel will be notified through the communication network to ensure that the fire extinguisher is always in good working condition.

[0099] Through the combination of sensor modules and data processing models, real-time monitoring of the transformer operating status and temperature prediction are achieved, and potential overheating risks are discovered in advance. The Transformer-based large model architecture and multi-head self-attention mechanism are used to analyze the factors affecting temperature from multiple angles to improve the accuracy and reliability of temperature prediction. When a fire or danger warning is detected, the fire extinguishing module can be quickly activated, and the non-pressure storage fire extinguisher sprays the fire extinguishing agent through high-pressure gas to quickly extinguish the fire source and prevent the accident from expanding. The self-check and fault handling module ensures the normal working state of the fire extinguisher, prevents fire extinguishing failure caused by equipment failure, and improves the system The reliability of the system is improved, edge computing provides real-time processing capabilities, and cloud computing provides data storage and model optimization. The two work together to improve the overall performance and intelligence level of the system. Fire extinguishing agents that are suitable for electrical fires and do not harm transformers, such as heptafluoropropane, are selected to avoid secondary damage to the equipment during the fire extinguishing process and ensure the integrity of the transformer. By setting multi-level temperature thresholds, the system can take different response measures according to the degree of risk, which not only ensures safety but also avoids unnecessary intervention. By collecting external data such as ambient temperature, humidity, and air pressure, the system can adapt to different environmental conditions and improve the accuracy of prediction and the adaptability of the system.

[0100] Example 2

[0101] like Figures 1 to 3 As shown, based on Example 1, Example 2 describes in detail the system's prediction and early warning mechanism, the improvement of the fire extinguishing module, the enhancement of the self-checking and fault handling functions, and the improvement of the external data integration and online learning capabilities.

[0102] The prediction and warning module sets four levels of temperature thresholds:

[0103] Normal level: The temperature is within the safe range and the system is monitoring normally.

[0104] Attention level: When the temperature approaches the lower limit of the set threshold, the system issues an attention warning signal to remind operation and maintenance personnel to strengthen monitoring.

[0105] Warning level: When the temperature exceeds the warning threshold, the system sends out a warning signal and automatically enters the fire extinguishing preparation state.

[0106] Danger level: The temperature far exceeds the threshold, the system sends out a danger warning signal and immediately activates the fire extinguishing module.

[0107] The transmission process of early warning signals includes:

[0108] Sound and light alarm: Sound and light alarms are used to alert on-site personnel.

[0109] Remote notification: Warning information is sent to operation and maintenance personnel and management center via SMS, email or dedicated App.

[0110] There are various options for fire extinguishing agents: inert gases, such as nitrogen and argon, are suitable for electrical fires and are non-corrosive to equipment; clean gases, such as heptafluoropropane and sulfur hexafluoride, have high fire extinguishing efficiency and little residue; dry powder fire extinguishing agents are special dry powders suitable for electrical fire extinguishing and have low fire extinguishing costs.

[0111] The startup process is controllable: according to the fire intensity and equipment conditions, the injection pressure of the fire extinguishing agent is adjusted to avoid mechanical damage to the equipment, and a reasonable injection duration is set to ensure the fire extinguishing effect while reducing the waste of fire extinguishing agent.

[0112] The system automatically triggers self-inspection at preset time intervals (such as weekly or monthly). Operation and maintenance personnel can manually trigger self-inspection remotely or on-site. When the system issues an early warning signal, self-inspection is automatically triggered to ensure that the fire extinguisher is in normal working condition.

[0113] Check the integrity of the power supply and starting circuit to ensure normal startup, detect the remaining amount and effectiveness of the fire extinguishing agent to prevent the fire extinguishing agent from being expired or insufficient, check the patency and sealing of the nozzle and pipeline to prevent blockage or leakage, check the system's communication module and control signal transmission status, and ensure that instructions can be issued accurately.

[0114] If any abnormality is detected, the system will immediately send out an audible and visual alarm signal and send the fault information to the remote monitoring center or the maintenance personnel's mobile device through the communication module, including the fault type, location and recommended treatment measures.

[0115] By accessing the meteorological data interface, weather data such as temperature, humidity, wind speed, rainfall and air pressure can be obtained in real time. Through the data interface with the power dispatching center, the load, power supply frequency and voltage fluctuation of the local power grid can be obtained.

[0116] The external data is fused with the data from the sensor module as the input of the model to improve the accuracy of temperature prediction.

[0117] The system has the ability to learn online and continuously adjust and optimize model parameters based on newly acquired data and operating conditions. For abnormal data or new operating modes, the model can learn and adjust in a timely manner to improve the ability to predict special situations.

[0118] Working process

[0119] The system sets multi-level temperature thresholds based on the historical operating data and current real-time data of the transformer, including normal level, attention level, warning level and danger level.

[0120] The data processing and model module uses the fused multi-source data to predict the temperature status of the transformer in real time. The prediction and early warning module compares the prediction results with the set temperature threshold to determine the current risk level.

[0121] According to different risk levels, the system takes corresponding measures: Normal level: The system continues to monitor normally and no special processing is required. Attention level: The system sends out an attention warning signal to remind the operation and maintenance personnel to pay attention and may increase the monitoring frequency. Warning level: The system sends out a warning warning signal, the fire extinguishing module enters the ready state, and the self-test and fault handling module starts self-test. Danger level: The system sends out a danger warning signal and immediately starts the fire extinguishing module to extinguish the fire.

[0122] When the fire extinguishing module receives the start command, it starts the fire extinguisher according to the preset parameters (injection pressure and time). The gas generator generates high-pressure gas, driving the fire extinguishing agent to spray through the nozzle to the overheated or fire area of ​​the transformer. During the fire extinguishing process, the system continuously monitors the temperature and fire situation of the transformer, and adjusts the fire extinguishing parameters or performs secondary fire extinguishing when necessary. After the fire extinguishing is completed, the system records the fire extinguishing process data and notifies the operation and maintenance personnel to conduct on-site inspection and equipment recovery.

[0123] In the early warning state or under regular triggering, the self-test and fault handling module starts the self-test process. According to the predetermined test content, it checks the key components and functions of the fire extinguisher and the system one by one. If the test passes, the system records the self-test results and maintains normal operation. If an abnormality is detected, the system immediately sends an alarm signal and sends detailed fault information to the remote monitoring center or maintenance personnel, and recommends taking appropriate repair or replacement measures.

[0124] The external data integration module obtains weather data and power grid operation data in real time. The data processing and model module integrates external data with sensor data as the input of the model. The online learning algorithm continuously adjusts the model parameters according to the new data to improve the accuracy of the prediction. The system regularly evaluates the performance of the model. If the prediction accuracy is found to have decreased, it may trigger the retraining or optimization of the model.

[0125] Through multi-level early warning, the system can take appropriate measures according to different risk levels to avoid overreaction or underreaction. At the attention and warning levels, the system has begun to strengthen monitoring and prepare for firefighting, which helps prevent the escalation of risks. Remote notification of early warning signals ensures that operation and maintenance personnel are aware of the equipment status in a timely manner to facilitate coordinated processing.

[0126] According to the actual application scenario, the most suitable fire extinguishing agent is selected to improve fire extinguishing efficiency and reduce costs. The adjustable injection pressure and time can avoid secondary damage to the transformer and surrounding equipment and extend the life of the equipment. The self-check mechanism ensures that the fire extinguishing system is always in good condition to avoid fire extinguishing failure due to equipment failure at critical moments. The real-time transmission of fault information enables maintenance personnel to take timely measures to reduce the impact of faults on system operation. By integrating weather and power grid data, the model can more comprehensively consider factors affecting temperature and make predictions more accurate. The online learning capability enables the model to adapt to environmental changes and new operating modes and maintain a high level of prediction performance. From data collection, prediction and warning to fire extinguishing, the system has achieved a high degree of automation and reduced human intervention. The detailed warning and fault information provided by the system provides strong support for operation and maintenance and management decisions.

[0127] By discovering potential problems in advance, reducing failures and downtime, lowering maintenance costs, and rationally arranging manpower and materials according to risk levels, we can avoid wasting resources.

[0128] Example 3

[0129] like Figures 1 to 3 As shown, this embodiment, based on Embodiment 1 and Embodiment 2, describes in detail the working method of the intelligent fire extinguishing system for transformers, specifically including steps S1 to S8, and aims to achieve real-time monitoring, temperature prediction, early warning and fire extinguishing of transformers through a systematic process.

[0130] Step S1: Data collection

[0131] Power sensors are installed at the input and output ends of the transformer to measure input power and output power, and calculate the power difference to evaluate the heat generation of the transformer. Temperature sensors are distributed in key parts of the transformer, such as windings, oil tanks and radiators, to obtain comprehensive temperature information. Voltage and current sensors monitor the voltage and current parameters of the transformer.

[0132] Environmental sensors collect data such as ambient temperature, humidity, and air pressure. External data integration module: obtains weather data (wind speed, rainfall, and atmospheric pressure) and grid operation data (load, power supply frequency, and voltage fluctuations).

[0133] The sensor collects data at a high frequency (e.g., once per second) to ensure the real-time nature of the data, and the data is transmitted to the data processing and model module via wired or wireless communication.

[0134] Step S2: Data preprocessing

[0135] Data cleaning: Remove outliers and noisy data, fill missing values, and use filtering algorithms such as Kalman filtering to smooth the data.

[0136] Data normalization: Convert data of different dimensions to the same scale range using Min-Max normalization or Z-score normalization method.

[0137] Feature extraction: Extract key features, such as power difference, temperature gradient, voltage fluctuation rate, etc., and generate feature vectors to facilitate model processing.

[0138] Data preprocessing is completed in real time in the edge computing module to ensure low latency, and data processing programs written in Python or C++ are used to improve efficiency.

[0139] Step S3: Temperature prediction

[0140] S3.1: Model input preparation

[0141] The feature vector preprocessed in step S2 is input into the data processing and model module. The input data includes the fusion of internal and external data, such as power difference, temperature, environmental parameters, weather data, grid load, etc. The data is organized in time series to form an input sequence, which is suitable for the time series model.

[0142] S3.2: Feature association analysis

[0143] Use as Figure 4 The formula shown uses the self-attention mechanism to calculate the correlation weights between input features and identify key factors that have a significant impact on transformer temperature, such as load fluctuations and ambient temperature changes.

[0144] S3.3: Multi-head self-attention mechanism processing

[0145] In the Transformer model, a multi-head self-attention mechanism is used to process features in parallel from different subspaces. Each attention head focuses on different feature combinations to capture complex nonlinear relationships and interactions. For example, one attention head may focus on the correlation between power difference and internal temperature, while another may focus on the impact of ambient humidity on heat dissipation efficiency. An appropriate number of heads (such as 8 heads) is set to balance the complexity and computational efficiency of the model.

[0146] S3.4: Deep feature extraction and representation

[0147] By stacking multiple layers of Transformer encoders, features are deeply extracted to obtain a high-dimensional representation of the transformer temperature state, capturing the characteristics of time series and spatial distribution. Residual connections and Layer Normalization are used to prevent gradient disappearance and stabilize the training process.

[0148] S3.5: Current temperature status prediction

[0149] The high-dimensional feature representation is input into the fully connected layer or regression layer, the temperature values ​​of the current key parts of the transformer are calculated, the overall temperature distribution of the transformer is output, and possible overheating areas are identified. The loss function uses the mean square error (MSE) to optimize the model parameters.

[0150] S3.6: Prediction of future temperature trends

[0151] Based on the historical data and time series patterns learned by the model, the temperature change trend within a specified time range in the future is predicted, the future temperature change curve is generated, and the possible temperature peak and change rate are predicted. The prediction time range can be set to the next 1 hour or longer to meet different application requirements.

[0152] S3.7: Risk Assessment and Anomaly Detection

[0153] Conduct risk assessment on the predicted temperature results, calculate the probability and possibility of temperature exceeding the limit, use the set temperature threshold to perform anomaly detection, identify potential overheating risks and abnormal temperature rise, and use probability models or threshold judgment methods to determine the risk level.

[0154] S3.8: Prediction result output

[0155] Integrate the current temperature status, future temperature trend, risk assessment and anomaly detection results, and output the integrated prediction results to the prediction and early warning module to provide a basis for subsequent early warning and processing. At the same time, transmit key prediction data to the edge computing and cloud computing collaborative module for data backup and model optimization, and use data compression and encryption technology to ensure the efficiency and security of data transmission.

[0156] Step S4: Edge and cloud computing collaboration

[0157] Perform temperature prediction and warning locally in real time, with a response time of less than milliseconds, temporarily store recent data to prevent data loss due to network interruptions, store and back up historical data for a long time, use high-performance computing resources in the cloud, regularly train and optimize models, synchronize data between the edge and the cloud regularly to ensure consistency, and send the cloud-optimized model to the edge computing module through the network to achieve iterative updates of the model. Use protocols such as MQTT or HTTP to achieve communication between the edge and the cloud, and use SSL / TLS encryption to ensure data transmission security.

[0158] Step S5: Early warning mechanism

[0159] Based on the predicted temperature and risk assessment results, the temperature is compared with the set multi-level temperature thresholds.

[0160] Warning signal issued:

[0161] Attention level: Issues an attention warning signal to remind operation and maintenance personnel to strengthen monitoring.

[0162] Warning level: Issue a warning signal, prepare to extinguish the fire, and start self-check.

[0163] Danger level: A danger warning signal is issued and the fire extinguishing module is activated to enter the preparatory working state.

[0164] The early warning signal is conveyed through sound and light alarms, text messages, emails, etc. The early warning information includes temperature data, risk level, recommended measures, etc.

[0165] Step S6: Infrared monitoring

[0166] The infrared camera obtains thermal imaging data from the outside of the transformer to monitor the temperature distribution. The visible light camera obtains visible light images to assist in judging the fire situation. The thermal imaging and visible light images are combined to detect whether there is a fire or abnormally high temperature area. The image data is processed by the edge computing module and image recognition algorithms, such as CNN, are used to identify flame characteristics.

[0167] Step S7: Self-check and troubleshooting

[0168] When the early warning signal is issued, the self-check and fault handling module automatically triggers the self-check mechanism of the fire extinguisher, detects the integrity of the power supply and starting circuit, detects the remaining amount and validity period of the fire extinguishing agent, detects whether the nozzle and pipeline are unobstructed, whether there is any blockage, and checks the electrical connection and signal transmission status of the system. If the fire extinguisher is detected to be invalid or abnormal, an alarm signal will be issued in time, and the fault information will be sent to the remote monitoring center or the maintenance personnel's mobile device through the communication module. The self-check process is controlled by the microcontroller with a short cycle and does not affect the normal operation of the system. The fault information includes the fault type, location, and recommended processing methods.

[0169] Step S8: Fire extinguishing

[0170] When the infrared monitoring module detects a fire or receives a temperature warning of a dangerous level, the fire extinguishing module activates the non-pressure storage fire extinguisher, and the gas generator produces high-pressure gas to drive the spray of fire extinguishing agent, which is suitable for electrical fires (such as heptafluoropropane, carbon dioxide) to cover the surface of the transformer and the internal overheated area. The system continuously monitors the temperature and fire conditions during the fire extinguishing process to ensure the fire extinguishing effect. After the fire extinguishing is completed, the system records the fire extinguishing data and notifies the operation and maintenance personnel to conduct on-site inspections and equipment recovery. The fire extinguishing agent spray pressure and time can be adjusted according to actual conditions to avoid secondary damage to the equipment. The fire extinguishing process is precisely controlled by a dedicated controller to ensure safety.

[0171] Through clear steps S1 to S8, the system realizes the automation of the whole process from data collection to fire extinguishing, with clear logic and efficient operation. By using the Transformer model and multi-head self-attention mechanism, the model can accurately predict the current and future temperature states and detect potential risks in advance. Through risk assessment and anomaly detection, the system can calculate the probability of temperature exceeding the limit in real time and take preventive measures in time to avoid accidents. Edge computing ensures the real-time and reliability of the system. Cloud computing provides powerful data storage and model optimization capabilities. The two work together to improve system performance. The multi-level early warning mechanism enables the system to take different measures according to the risk level, from strengthening monitoring to starting fire extinguishing to ensure safety. The self-check mechanism ensures the reliability of key equipment. The fault handling process detects and handles anomalies in time, reducing the risk of system operation. The application of non-pressure storage fire extinguishers, the fire extinguishing agent is suitable for electrical fires, starts quickly, has good fire extinguishing effect, and does not damage the equipment. Through modular design, the system can expand and upgrade functions as needed, such as adding new sensor types or optimizing model algorithms. Encrypted communication and data compression technology are used to ensure the security and integrity of data during transmission. The high degree of automation reduces manual intervention and maintenance costs and avoids economic losses caused by failures.

[0172] In view of current practical needs, the above-mentioned implementation mode adopted in this application is not limited to the scope of protection. Various changes made within the knowledge scope of technical personnel in this field without departing from the concept of this application still fall within the scope of protection of the present invention.

Claims

1. An intelligent fire extinguishing system for transformers, characterized in that: It includes sensor module, data processing and model module, edge computing and cloud computing collaboration module, prediction and early warning module, infrared monitoring module, fire extinguishing module and self-test and fault handling module; The sensor module is used to collect internal and external data of the transformer, including but not limited to power change, input power and output power difference, transformer internal temperature, voltage, current, ambient temperature, humidity and air pressure; The data processing and model module is connected to the sensor module to pre-process the collected data, including data cleaning, normalization and feature extraction. Based on the Transformer large model architecture and self-attention mechanism, the data is converted into a vector representation that can be understood by the model, and the multi-head self-attention mechanism is used to analyze the various factors affecting the transformer temperature, and the current temperature state and future temperature change trend of the transformer are predicted in real time; The edge computing and cloud computing collaboration module is connected to the data processing and model module. The edge computing module is used for real-time processing and response, and performs temperature prediction and early warning. The edge computing and cloud computing collaboration module includes an edge computing module and a cloud computing module. The cloud computing module is used for long-term data storage, backup and model training optimization. The edge computing module and the cloud computing module work together through an efficient data communication protocol; The prediction and warning module is connected to the data processing and model module and the edge computing and cloud computing collaboration module. According to the temperature prediction result, when the predicted temperature may exceed the set threshold, a multi-level warning signal is issued, and the fire extinguishing module is activated to enter a preparatory working state; The infrared monitoring module includes an infrared camera and a visible light camera, which monitors the temperature changes and fire conditions outside the transformer in real time, and detects whether a fire or abnormally high temperature area occurs by combining thermal imaging and visible light images; The fire extinguishing module includes a non-pressure storage type fire extinguisher, which is connected to the infrared monitoring module and the prediction and warning module. When the infrared monitoring module detects a fire or receives a temperature warning of a dangerous level, the fire extinguisher is activated to extinguish the fire. The fire extinguisher generates high-pressure gas through an internal gas generator to spray a fire extinguishing agent, which is suitable for electrical fires and does not damage transformers; The self-check and fault handling module is connected to the fire extinguishing module. When an early warning signal is issued, the self-check mechanism of the fire extinguisher is triggered to detect the gas generator status, fire extinguishing agent capacity, nozzle patency and electrical connection integrity of the fire extinguisher to ensure that the fire extinguisher is in normal working condition. If the fire extinguisher is detected to be invalid or abnormal, an alarm signal is issued in time to notify the maintenance personnel.

2. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The sensor module includes a power sensor, which is used to measure the input power and output power of the transformer and calculate the power difference to evaluate the heat generation of the transformer; the temperature sensors are distributed in multiple key parts of the transformer, including windings, oil tanks and radiators, to obtain comprehensive temperature information.

3. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The data processing and model module adopts a Transformer-based large model architecture and uses a multi-head self-attention mechanism to analyze the factors affecting transformer temperature from multiple angles such as load changes, ambient temperature, heat dissipation capacity, weather conditions, and grid load data, thereby improving the accuracy and reliability of temperature prediction.

4. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The edge computing module in the edge computing and cloud computing collaborative module includes a high-performance processor and a storage unit, which can execute complex prediction and early warning algorithms locally in real time; the cloud computing module is based on a cloud server and uses big data analysis and machine learning technology to conduct in-depth mining of historical data and continuously optimize the prediction model.

5. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The prediction and early warning module sets multiple levels of temperature thresholds, including normal level, attention level, warning level and danger level. When the predicted temperature exceeds different thresholds, early warning signals of corresponding levels are issued respectively. At the normal level, monitoring is maintained; at the attention level, monitoring is strengthened; at the warning level, fire extinguishing is prepared; and at the danger level, the fire extinguishing module is immediately activated.

6. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The fire extinguishing agent of the non-pressure storage fire extinguisher in the fire extinguishing module is an inert gas, clean gas or dry powder suitable for electrical fire extinguishing, including but not limited to heptafluoropropane, sulfur hexafluoride or carbon dioxide. The starting process of the fire extinguisher is controlled, and the injection pressure and time are adjustable to avoid secondary damage to the transformer and surrounding equipment.

7. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The self-test and fault handling module can trigger the self-test of the fire extinguisher on a regular, manual and early warning basis, and the test contents include but are not limited to: the integrity of the power supply and starting circuit of the gas generator; the remaining amount and effectiveness of the fire extinguishing agent; the patency and sealing of the nozzle and pipeline; the communication and control signal transmission status of the system; if any abnormality is detected, an alarm signal is immediately issued, and the fault information is sent to the remote monitoring center or the mobile device of the maintenance personnel through the communication module.

8. The intelligent fire extinguishing system for transformer according to claim 1, characterized in that: The external data integration module obtains real-time weather data, including temperature, humidity, wind speed, rainfall and air pressure, as well as the load, power supply frequency and voltage fluctuation of the local power grid, and provides these data together with the data of the sensor module to the data processing and model module to optimize the temperature prediction model; the system has online learning capabilities and can continuously adjust and optimize model parameters according to newly acquired data and operating conditions to improve the accuracy of prediction and warning.

9. An intelligent fire extinguishing system for transformer according to any one of claims 1 to 8, characterized in that: Its working method includes the following steps: S1: Data collection, collecting internal and external data of the transformer through the sensor module, the data including power change, input power and output power difference, transformer internal temperature, voltage, current, ambient temperature, humidity and air pressure; S2: Data preprocessing: In the data processing and model module, the collected data is cleaned, normalized, and features are extracted to convert the data into a vector representation that can be understood by the model; S3: Temperature prediction S3.1: Model input preparation Input the vectorized data preprocessed in step S2 into the data processing and model module, the data including but not limited to: power change data of the transformer; difference between input power and output power; temperature data of internal temperature, including the temperature of windings, oil tank and radiator; voltage and current data; environmental parameters of ambient temperature, humidity and air pressure; weather condition data, such as wind speed, rainfall and atmospheric pressure; local power grid load data and voltage fluctuation; S3.2: Feature association analysis S3.2.1: Use the self-attention mechanism to perform feature correlation analysis on the input data; S3.2.2: Calculate the correlation weights between the features and identify the key factors that have a significant impact on the transformer temperature, such as load fluctuations and ambient temperature changes; S3.3: Multi-head self-attention mechanism processing S3.3.1: In the Transformer large model architecture, a multi-head self-attention mechanism is used to process features from different subspaces in parallel; S3.3.2: Each attention head focuses on different feature combinations to capture complex nonlinear relationships and interactions, such as the correlation between power difference and internal temperature, and the impact of ambient humidity on heat dissipation efficiency; S3.4: Deep feature extraction and representation S3.4.1: Deeply extract features processed by the multi-head self-attention mechanism by stacking multiple layers of encoders; S3.4.2: Generate a high-dimensional representation of the transformer temperature state that captures the characteristics of the time series and spatial distribution; S3.5: Current temperature status prediction S3.5.1: Input the high-dimensional feature representation into the fully connected layer or regression layer to calculate the current temperature values ​​of each key part of the transformer; S3.5.2: Temperature distribution of the output transformer as a whole, identifying possible overheating areas; S3.6: Prediction of future temperature trends S3.6.1: Based on the historical data and time series patterns learned by the model, predict the temperature change trend within a specified time range in the future; S3.6.2: Generate future temperature change curves to predict possible temperature peaks and rates of change; S3.7: Risk Assessment and Anomaly Detection S3.7.1: Perform a risk assessment on the predicted temperature results and calculate the probability and likelihood of temperature exceedances; S3.7.2: Use set temperature thresholds to perform anomaly detection and identify potential overheating risks and abnormal temperature rise conditions; S3.8: Prediction result output S3.8.1: Integrate current temperature status, future temperature trends, risk assessment, and anomaly detection results; S3.8.2: Output the integrated forecast results to the forecast and warning module to provide a basis for subsequent warning and processing; S3.8.3: At the same time, key prediction data is transmitted to the edge computing and cloud computing collaboration modules for data backup and model optimization; S4: Edge and cloud computing work together to perform temperature prediction and early warning in real time through edge computing modules and cloud computing modules, and use cloud computing modules to perform long-term data storage, backup, and model training optimization; S5: Early warning mechanism: According to the temperature prediction results, when the predicted temperature may exceed the set multi-level threshold, the prediction and early warning module sends out an early warning signal of the corresponding level and activates the fire extinguishing module to enter the preparatory working state; S6: Infrared monitoring: The infrared camera and visible light camera of the infrared monitoring module monitor the temperature changes and fire conditions outside the transformer in real time to detect whether there is a fire or abnormally high temperature area; S7: Self-check and fault handling. When the warning signal is issued, the self-check and fault handling module triggers the self-check mechanism of the fire extinguisher to detect the gas generator status, fire extinguishing agent capacity, nozzle patency and electrical connection integrity. If the fire extinguisher is detected to be invalid or abnormal, an alarm signal is issued in time and the maintenance personnel are notified to handle it; S8: Fire extinguishing process. When the infrared monitoring module detects a fire or receives a temperature warning of a dangerous level, the fire extinguishing module activates the non-pressure storage fire extinguisher. The internal gas generator produces high-pressure gas and sprays a fire extinguishing agent suitable for electrical fires to extinguish the transformer.

Images