A high-altitude energy storage power station fire warning method and system

Abstract

The application provides a high-altitude energy storage power station fire warning method and system, belonging to the technical field of energy storage power station fire warning, and comprising the following processing steps: data acquisition: a thermal imaging sensor, a visible light camera and an air pressure sensor perform original data stream acquisition in a preset timestamp synchronization mechanism, and three-mode data is input into a high-altitude dynamic registration and compensation module and the like; through dynamic compensation driven by environmental parameters, three-layer progressive deep fusion, four-thread cooperative verification, low-pressure heat spread deduction, performance closed-loop evaluation and knowledge graph self-evolution, a full-link intelligent closed loop of "perception-cognition-decision-action-evolution" is constructed, and the system has the advantages of strong adaptability, excellent response timeliness, high resource utilization rate and strong system evolution capability.

Description

Technical Field

[0001] This application relates to the field of fire early warning technology for energy storage power stations, and in particular to a fire early warning method and system for high-altitude energy storage power stations. Background Technology

[0002] With the large-scale deployment of electrochemical energy storage power stations in high-altitude areas (above 3000 meters), their fire safety protection faces the dual challenges of adaptability to extreme environments and the complexity of battery thermal runaway mechanisms. Traditional fire early warning systems for energy storage power stations mostly use smoke and heat detectors linked to fixed threshold alarms, relying on manual inspections or simple video monitoring for verification. In recent years, some improved solutions have attempted to introduce thermal imaging technology for temperature field monitoring or integrate visible light video for image recognition, which has improved fire identification capabilities to some extent. However, these existing systems still reveal the following systemic defects in the special environment of high-altitude low air pressure, strong ultraviolet radiation, and large temperature fluctuations: The current detection methods are limited to a single dimension and cannot identify early thermal runaway micro-hot spots in energy storage power stations. Relying on absolute temperature thresholds or smoke concentration to determine the fire situation, existing systems can only respond to open flames or large-scale thermal runaways (typically with temperature rises exceeding 20°C). They are insensitive to micro-hot spots (initial anomalies with temperature rises of only 3-5°C) caused by internal short circuits in battery modules or poor terminal contact. Furthermore, existing technologies lack ultraviolet spectrum detection, failing to capture early arc discharge characteristics (150-280nm ultraviolet pulses) in thermal runaway, thus missing the window for identifying early electrical faults. They also lack visible light casing deformation monitoring, making it difficult to identify structural failure signs such as battery bulging and cracked explosion-proof valves. This single-dimensional detection results in a delayed system response, preventing intervention before the thermal runaway chain reaction begins.

[0003] The lack of a dynamic compensation mechanism for high-altitude environmental parameters leads to inaccurate early warning models. At high altitudes, air pressure is 30%-50% lower than standard atmospheric pressure, significantly altering the optical focal length and thermal radiation transmission path of thermal imagers. This results in pixel-level misalignment errors in traditional fixed-parameter registration methods, causing a continuous degradation in the spatial alignment accuracy between thermal imaging and visible light video. Existing systems do not perform real-time monitoring and dynamic registration compensation of air pressure parameters, resulting in micro-hotspot location deviations of tens of pixels, affecting the accuracy of subsequent targeted fire suppression. Simultaneously, the reduced efficiency of thermal convection under low-pressure environments causes traditional thermal spread prediction algorithms based on atmospheric pressure models to severely overestimate the fire spread rate, leading to inflated early warning levels and wasted emergency response resources.

[0004] The early warning response fails to consider the thermal spread characteristics of low-pressure environments, leading to distorted situation predictions. The existing system only outputs a fixed-level alarm after confirming a fire, without coupling the energy storage power station's battery cluster layout model, cabin ventilation path data, and low-pressure thermal diffusion correction model for simulation. This makes it impossible to generate fire spread timing predictions that accurately reflect the realities of the plateau environment. Consequently, emergency plans are inaccurate, and personnel evacuation route planning fails to consider the differences in smoke deposition characteristics caused by low pressure, posing safety hazards.

[0005] The system lacks closed-loop assessment and reignition detection capabilities for fire extinguishing effectiveness. Traditional systems, after activating the extinguishing device, lack real-time monitoring of the extinguishing agent concentration field within the chamber, the fire source energy decay trend, and whether arc characteristics have disappeared. This makes it impossible to determine whether perfluorohexanone effectively covers micro-hotspot areas. This open-loop approach of "releasing without regard to effect" fails to identify extinguishing failures or pre-ignition signs of thermal runaway due to insufficient extinguishing agent diffusion in low-pressure environments (secondary temperature gradient rise), missing the window for secondary intervention. Furthermore, failure data cannot be fed back to correct the judgment threshold of the front-end identification engine, preventing the system from learning from errors and resulting in performance degradation.

[0006] The system lacks self-learning and knowledge evolution mechanisms, making it unable to adapt to diverse scenarios. High-altitude energy storage power stations exhibit significant variations in battery type, cabin structure, altitude, and ventilation layout. The existing system lacks structured storage for key data such as the causes of micro-hotspots, environmental parameters, release efficiency, and final results of each warning and fire suppression event, failing to form a reusable knowledge graph. The system cannot adaptively optimize recognition sensitivity, registration compensation coefficients, and fire extinguishing agent release strategies through accumulated operational data. This necessitates repeated on-site debugging in new scenarios, with cycles lasting several weeks, and accuracy is difficult to continuously improve after long-term operation.

[0007] The lack of consideration for strong ultraviolet (UV) light interference leads to a surge in false alarm rates in the visible light channel. UV intensity in high-altitude areas is 2-3 times stronger than in plains, causing UV fluorescence interference and pixel overexposure on the visible light camera lens surface, severely impacting the accuracy of battery casing deformation detection. The existing system does not perform real-time monitoring and adaptive suppression of UV intensity, nor does it dynamically reduce the weight of the visible light channel in the fusion decision-making process. This results in a significant increase in false alarm rates during periods of strong sunlight, severely limiting system availability.

[0008] In summary, existing fire early warning systems for energy storage power stations have structural defects in key areas such as micro-hotspot detection, environmental compensation, situation prediction, performance evaluation, and self-learning evolution. In particular, they cannot adapt to special operating conditions such as high altitude, low air pressure, and strong ultraviolet interference. There is an urgent need for a systematic solution that can achieve multimodal collaborative detection, high-altitude dynamic registration compensation, low-pressure thermal spread simulation, closed-loop feedback of fire extinguishing performance, and knowledge-driven evolution. Summary of the Invention

[0009] To address the aforementioned issues, this application provides a fire early warning method and system for high-altitude energy storage power stations. Through dynamic compensation driven by environmental parameters, three-layer progressive deep fusion, four-thread collaborative verification, low-pressure thermal propagation simulation, performance closed-loop evaluation, and knowledge graph self-evolution, a full-link intelligent closed loop of "perception-cognition-decision-action-evolution" is constructed. It has strong adaptability, excellent response timeliness, high resource utilization, and strong system evolution capability.

[0010] This application provides a fire early warning method for high-altitude energy storage power stations, including the following processing steps: Data acquisition: The thermal imaging sensor, visible light camera and barometric pressure sensor acquire raw data streams using a preset timestamp synchronization mechanism, and input the three-modal data into the high-altitude dynamic registration and compensation module; Dynamic registration and compensation: The high-altitude dynamic registration and compensation module acquires barometric pressure sensor data in real time, calculates the offset of the thermal imager's optical parameters corresponding to the current barometric pressure value, and generates a barometric pressure compensation transformation matrix. Simultaneously, it acquires the temperature and humidity deviation, illumination fluctuation, and ultraviolet light intensity values ​​output by the environmental parameter adaptive unit. When the ultraviolet light intensity value exceeds a preset interference threshold, the thermal imaging mode is locked as the registration reference, and the stable heat source contour in the thermal image is extracted as the registration anchor point. When the ultraviolet light intensity value is lower than the interference threshold and the barometric pressure value is within a stable range, the visible light mode is locked as the registration reference, and the structural edge features in the visible light image are extracted as the registration anchor point. Based on the selected registration anchor point and the barometric pressure compensation transformation matrix, the spatial transformation residual of the current frame relative to the reference template is calculated, generating high-altitude adaptive registration parameters to complete pixel-level spatial alignment of the three-modal data. The registered three-modal data is then input into the high-altitude three-layer progressive fusion module. Three-layer progressive fusion: The high-altitude three-layer progressive fusion module performs progressive processing: First, at the pixel level, complementary energy distribution enhancement is performed on thermal imaging and visible light images, while high-frequency noise components in the visible light channel are dynamically suppressed according to the ultraviolet light intensity value, generating a dual-channel fused image; Second, at the feature level, thermal gradient variation features and visual semantic association features of the fused image are extracted, and a pressure correction factor is superimposed to normalize the thermal gradient amplitude, constructing a three-modal joint feature vector; Finally, at the decision level, the joint feature vector is dynamically weighted, where the lower the pressure value, the more adaptively the weight of the thermal imaging channel is increased, and the higher the ultraviolet light intensity value, the more adaptively the weight of the visible light channel is decreased, generating a fusion decision map with high-altitude thermal anomaly sensitivity and visual interpretability, which is output to the multi-dimensional thermal runaway verification engine; Thread Verification: After receiving the fused decision graph, the multi-dimensional thermal runaway verification engine starts four verification threads in parallel: The first thread extracts the micro-hotspot connected regions in the fused decision graph and calculates the thermal diffusion compliance and temperature rise rate index of the region; the second thread performs arc flicker spectrum analysis and battery casing deformation trajectory fitting on the corresponding region in the visible light channel; the third thread calls historical thermal runaway precursor samples of the same scenario based on the energy storage power station's dedicated knowledge graph for pattern similarity comparison; the fourth thread calls the low-pressure thermal radiation propagation correction model according to the current air pressure value to perform environmental compensation calculation on the micro-hotspot temperature field; if and only if the verification results of all four threads pass the preset logic threshold and meet the spatiotemporal consistency constraints, the target is marked as a credible thermal runaway source, and a thermal runaway confidence rating is generated and output to the graded early warning and plateau situation prediction module; if any thread verification fails, the abnormal features are fed back to the high-altitude dynamic registration and compensation module to trigger registration parameter correction; Tiered early warning and plateau situation prediction: The tiered early warning and plateau situation prediction module, based on the thermal runaway confidence rating and target spatial coordinates, combined with the thermal barrier structure constraints and cabin ventilation path data in the battery cluster layout model of the energy storage power station, calls the thermal runaway propagation simulation unit under low pressure environment to generate fire boundary projection results for future periods, and matches them with the response level in the plateau emergency plan library, and sends a tiered early warning instruction containing target location, hazard level, and spread trend to the emergency linkage and plateau effectiveness assessment module; Emergency Response and High-Altitude Performance Assessment: Upon receiving a graded early warning command, the emergency response and high-altitude performance assessment module executes progressive responses according to the response level: It prioritizes activating the battery cluster fire suppression device associated with the field of view for directional spraying; simultaneously triggers audible and visual alarms and dynamic planning of evacuation routes for power station personnel; receives real-time feedback pressure and flow data from the fire suppression device, combines it with the energy attenuation curve of the fire source area from the fusion decision map, and uses the air pressure correction factor to assess fire suppression performance; when the temperature gradient in the fire source area continuously decreases and the visual smoke concentration is below a preset threshold, it is determined to be an initial fire extinguishing and enters a continuous monitoring state; if the temperature gradient rises again or the target area splits and spreads, a secondary response upgrade is initiated, and real-time response data is transmitted back to the energy storage power station's dedicated knowledge graph for case updates; Adaptive closed-loop adjustment: During system operation, the environmental parameter adaptive unit continuously monitors the rate of change of air pressure, the rate of change of ultraviolet light intensity, and the rate of change of temperature and humidity. When any rate of change exceeds the preset high-altitude stable range, the recalibration process of the high-altitude dynamic registration and compensation module is automatically triggered, realizing minute-level adaptive closed-loop adjustment of system parameters in high-altitude environments.

[0011] In some embodiments, the processing steps of the high-altitude dynamic registration and compensation module specifically include: Establish initial calibration matrices for the external and internal parameters of the thermal imaging sensor, visible light camera, and barometric pressure sensor, and store them as a plateau reference registration template; While acquiring the current frame data in real time, the barometric pressure sensor data is acquired simultaneously, the deviation ratio between the current barometric pressure value and the standard atmospheric pressure is calculated, and converted into the focal length compensation coefficient of the thermal imager's optical lens. When the ultraviolet light intensity value exceeds the preset interference threshold, the thermal imaging mode is locked as the registration reference, the stable heat source contour in the thermal imaging is extracted as the registration anchor point, and the ultraviolet noise interference of the visible light channel is ignored. When the ultraviolet light intensity value is lower than the interference threshold and the air pressure deviation ratio is less than the preset fluctuation range, the visible light mode is locked as the registration reference, and the edge features of the battery cluster structure in the visible light image are extracted as the registration anchor point. By performing matrix coupling operations on the focal length compensation coefficient and the spatial transformation residual calculated based on the registration anchor point, high-altitude adaptive registration parameters are generated, and sub-pixel-level spatial alignment is achieved.

[0012] In some embodiments, the collaborative logical relationship of the four verification threads of the multi-dimensional thermal runaway verification engine is as follows: If the temperature rise rate index of the micro-hotspot area extracted by the first thread is lower than the threshold, the target is directly identified as a regular heat source interference and the subsequent threads are terminated. If the rate of temperature rise exceeds the threshold, the second thread is activated to analyze arc flash and casing deformation. When periodic arc flash or casing bulging deformation trajectory is detected in the visible light area, it is determined to be a precursor to battery thermal runaway, and the third thread is entered for verification. The third thread performs a similarity search between the feature vector of the target area and historical thermal runaway cases in the knowledge graph dedicated to energy storage power stations. If the similarity is higher than the warning value, it is marked as a suspected thermal runaway and the confidence rating is lowered, while triggering a manual review request; if the similarity is lower than the warning value, it is confirmed as a credible thermal runaway mode and the fourth thread is activated. The fourth thread calls the low-pressure thermal radiation propagation correction model based on the current air pressure value to perform environmental compensation calculations on the temperature field of the micro-hot spot. If the temperature value after compensation still exceeds the critical threshold for thermal runaway, a high confidence rating is output; otherwise, it is downgraded to a low confidence warning.

[0013] In some embodiments, the low-pressure thermal runaway propagation simulation unit of the graded early warning and plateau situation prediction module processes the following steps: Obtain the thermal conductivity of the thermal barrier material and the structural spacing parameters of each battery module in the battery cluster layout model; Acquire cabin ventilation path data, including air inlet location, exhaust fan speed, and airflow direction vector; Based on the current air pressure value, consult the table of low-pressure thermal convection efficiency attenuation coefficients and correct the convective heat transfer terms in the standard thermal diffusion model. The coordinates of the thermal runaway source, the modified thermal diffusion model, the battery cluster layout constraints, and the ventilation path data are input into the time series simulation engine to generate the boundary evolution results of the fire temperature field for future periods. The temperature field boundary evolution results are compared with the critical temperature of battery cluster combustion and explosion. When the predicted boundary touches the adjacent battery module, the warning level is automatically upgraded to the emergency response level.

[0014] In some embodiments, the barometric pressure correction factor evaluation step of the emergency response and high-altitude performance evaluation module is as follows: After the fire suppression device is activated, the average energy value and energy distribution dispersion of the fire source area in the fusion decision map are collected at fixed time intervals to construct an energy decay time series. Simultaneously collect real-time pressure and flow values ​​of the fire suppression device to construct an efficiency output time series; Based on the current air pressure value, the correction coefficient for the atomization efficiency of the extinguishing agent under low air pressure is queried, and the time series of performance output is normalized. The normalized performance output time series and energy decay time series are input into the trend coupling analysis unit: if the two show a positive correlation and the energy distribution dispersion continues to decrease, it is determined to be effective suppression and the current response level is maintained; if the two show a negative correlation or no correlation and the energy distribution dispersion increases, it is determined to be suppression failure, the response level is automatically upgraded and the fire suppression strategy is switched.

[0015] In some embodiments, the continuous updating mechanism of the energy storage power station-specific knowledge graph is as follows: The output results of each multi-dimensional thermal runaway verification engine, the handling data of the emergency response and high-altitude performance assessment module, and the final manual confirmation label are packaged into structured case data, and the current air pressure value and ultraviolet light intensity value are labeled as environmental context. The system performs battery module model clustering and thermal runaway mode classification on structured case data, automatically generating new thermal runaway precursor mode nodes or low-pressure environment interference nodes, and establishing causal relationship edges between nodes and related attributes of air pressure correction coefficients. When the number of knowledge graph nodes increases by a preset threshold, the system triggers the high-altitude model parameter adaptive optimization process, mapping the updated graph relationships to the logical threshold parameters of the multi-dimensional thermal runaway verification engine and the air pressure compensation coefficient of the high-altitude dynamic registration and compensation module, thereby realizing the closed-loop iteration of the system's discrimination capability in high-altitude scenarios.

[0016] This invention provides a fire early warning system for high-altitude energy storage power stations, comprising: The acquisition module, including a thermal imaging sensor, a visible light camera, and a barometric pressure sensor, is configured as a raw data stream acquisition device with timestamp synchronization. The high-altitude dynamic registration and compensation module includes a barometric compensation transformation unit, an ultraviolet interference determination unit, a registration reference selection unit, and a spatial alignment calculation unit, which are used to receive three-mode data and output high-altitude adaptive registration parameters. The high-altitude three-layer progressive fusion module includes a pixel-level noise suppression unit, a feature-level air pressure normalization unit, and a decision-level dynamic weight allocation unit, which are used to generate a fusion decision map. The multi-dimensional thermal runaway verification engine includes a micro hotspot analysis thread, an arc deformation analysis thread, a knowledge graph comparison thread, and a low-pressure correction thread, as well as a logic threshold unit for comprehensive judgment, which is used to output the thermal runaway confidence rating. The graded early warning and plateau situation prediction module includes a low-pressure thermal runaway propagation simulation unit and a plateau emergency plan matching unit, which are used to generate graded early warning instructions; The emergency response and high-altitude performance evaluation module includes a battery cluster fire suppression device control unit, a pressure correction performance evaluation unit, and a secondary response upgrade unit, which are used to perform progressive handling and generate feedback data. A dedicated knowledge graph library for energy storage power stations is bidirectionally connected to the multi-dimensional thermal runaway verification engine, emergency response and high-altitude performance evaluation module. It is used to store historical thermal runaway cases, low-pressure interference modes and handling data, and receive data for dynamic updates. The environmental parameter adaptive unit includes a barometric pressure sensor, an ultraviolet light sensor, a temperature and humidity sensor, and a rate of change determination subunit, which is used to monitor high-altitude environmental parameters in real time and trigger the system recalibration process. Specifically, the output of the environmental parameter adaptive unit is connected to the input of the high-altitude dynamic registration and compensation module, the output of the energy storage power station dedicated knowledge graph library is connected to the input of the multi-dimensional thermal runaway verification engine, and the feedback of the emergency linkage and plateau performance evaluation module is connected to the input of the energy storage power station dedicated knowledge graph library.

[0017] In some embodiments, the air pressure compensation transformation unit of the high-altitude dynamic registration and compensation module includes: The air pressure deviation calculation subunit is used to compare the real-time air pressure value with the standard atmospheric pressure to generate a deviation ratio; the focal length compensation coefficient query subunit is used to retrieve a preset air pressure-focal length offset mapping table based on the deviation ratio. The matrix coupling operation subunit is used to couple the focal length compensation coefficient with the spatial transformation residual to generate high-altitude adaptive registration parameters.

[0018] In some embodiments, the barometric pressure correction performance evaluation unit of the emergency response and high-altitude performance evaluation module includes: The energy decay monitoring subunit is used to extract the average energy value and distribution dispersion of the fire source area from the fusion decision map. The performance output monitoring subunit is used to collect real-time pressure and flow data of the fire suppression device; The low-pressure atomization correction subunit is used to normalize the performance output data based on the current air pressure value. The trend coupling analysis subunit is used to compare the correlation between the energy decay time series and the normalized performance output series, and output the judgment result of whether the suppression is effective or ineffective.

[0019] In some embodiments, it further includes: Power station-level edge computing nodes are deployed locally in each energy storage power station to perform real-time calculations for the high-altitude dynamic registration and compensation module, the high-altitude three-layer progressive fusion module, and the emergency response and plateau performance assessment module. The regional cloud analysis platform is connected to each power station-level edge computing node through a high-bandwidth, low-latency network. It is used to perform complex calculations and data storage for the multi-dimensional thermal runaway verification engine, the graded early warning and plateau situation prediction module, and the energy storage power station-specific knowledge graph library. The edge computing node and the cloud analysis platform exchange fusion decision graphs and thermal runaway verification results through a lightweight data transmission protocol, realizing closed-loop processing of edge-cloud collaboration in high-altitude energy storage power station scenarios.

[0020] Compared with the prior art, the present invention has the following beneficial effects: Through dynamic compensation driven by environmental parameters, three-layer progressive deep fusion, four-thread collaborative verification, low-pressure thermal spread extrapolation, performance closed-loop evaluation, and knowledge graph self-evolution, a full-link intelligent closed loop of "perception-cognition-decision-action-evolution" is constructed. It has strong adaptability, excellent response timeliness, high resource utilization, and strong system evolution capability. Attached Figure Description

[0021] The embodiments of the present invention will be further described below with reference to the accompanying drawings: Figure 1 A schematic diagram illustrating the implementation process of a fire early warning method for a high-altitude energy storage power station provided in this application embodiment; Figure 2 A schematic diagram of a fire early warning system for a high-altitude energy storage power station provided in this application embodiment; Figure 3 This is a schematic diagram of the composition structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0024] If the application documents contain similar descriptions such as "first, second, third", the following explanation shall be added: In the following description, the terms "first, second, third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0026] Based on the problems existing in related technologies, this application provides a fire early warning method for high-altitude energy storage power stations. The executing entity of this early warning method can be an electronic device. The electronic device can be various types of terminals such as laptops, tablets, desktop computers, set-top boxes, and mobile devices (e.g., mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable gaming devices), or it can be implemented as a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0027] In some embodiments, the functions implemented by the early warning method provided in this application can be achieved by the processor of an electronic device calling program code, wherein the program code can be stored in a computer storage medium.

[0028] This application provides a fire early warning method for high-altitude energy storage power stations. Figure 1 This application provides a schematic diagram illustrating the implementation process of a fire early warning method for high-altitude energy storage power stations, as shown in the embodiments below. Figure 1 As shown, the processing steps include the following: Step S1: Data Acquisition: The thermal imaging sensor, visible light camera and barometric pressure sensor acquire raw data streams using a preset timestamp synchronization mechanism, and input the three-modal data into the high-altitude dynamic registration and compensation module. Step S2: Dynamic Registration and Compensation: The high-altitude dynamic registration and compensation module acquires barometric pressure sensor data in real time, calculates the offset of the thermal imager's optical parameters corresponding to the current barometric pressure value, and generates a barometric pressure compensation transformation matrix. Simultaneously, it acquires the temperature and humidity deviation, illumination fluctuation, and ultraviolet light intensity values ​​output by the environmental parameter adaptive unit. When the ultraviolet light intensity value exceeds a preset interference threshold, the thermal imaging mode is locked as the registration reference, and the stable heat source contour in the thermal image is extracted as the registration anchor point. When the ultraviolet light intensity value is lower than the interference threshold and the barometric pressure value is within a stable range, the visible light mode is locked as the registration reference, and the structural edge features in the visible light image are extracted as the registration anchor point. Based on the selected registration anchor point and the barometric pressure compensation transformation matrix, the spatial transformation residual of the current frame relative to the reference template is calculated, generating high-altitude adaptive registration parameters to complete pixel-level spatial alignment of the three-modal data. The registered three-modal data is then input into the high-altitude three-layer progressive fusion module. In this invention, because the air pressure at high altitudes is 30%-50% lower than standard atmospheric pressure, it causes focal length shifts in the optical lens of thermal imagers and changes in the thermal radiation transmission path. Traditional fixed-parameter registration will produce pixel-level misalignment errors. This invention innovatively converts air pressure values ​​into focal length compensation coefficients, dynamically generating an air pressure compensation transformation matrix to achieve sub-pixel-level spatial alignment. Simultaneously, it intelligently switches the registration reference mode based on ultraviolet light intensity: when ultraviolet interference is strong, it locks the thermal imaging mode to avoid overexposure of the visible light channel; when the environment is stable, it locks the visible light mode, utilizing structural edge features to improve registration accuracy. This mechanism fundamentally solves the problem of thermal imaging and visible light video registration drift caused by high-altitude environments, controlling the spatial alignment error within three pixels. This provides a precise coordinate consistency basis for subsequent fusion analysis and avoids false negatives and missed positives caused by inaccurate registration.

[0029] In some embodiments, the processing steps of the high-altitude dynamic registration and compensation module specifically include: Step S21: Establish the initial calibration matrix of external and internal parameters of the thermal imaging sensor, visible light camera and barometric pressure sensor, and store it as a plateau reference registration template; Step S22: While acquiring the current frame data in real time, simultaneously acquire the barometric pressure sensor data, calculate the deviation ratio between the current barometric pressure value and the standard atmospheric pressure, and convert it into the focal length compensation coefficient of the thermal imager's optical lens. Step S23: When the ultraviolet light intensity value exceeds the preset interference threshold, lock the thermal imaging mode as the registration reference, extract the stable heat source contour in the thermal imaging as the registration anchor point, and ignore the ultraviolet noise interference of the visible light channel. Step S24: When the ultraviolet light intensity value is lower than the interference threshold and the air pressure deviation ratio is less than the preset fluctuation range, lock the visible light mode as the registration reference and extract the edge features of the battery cluster structure in the visible light image as the registration anchor point. Step S25: Perform matrix coupling operation between the focal length compensation coefficient and the spatial transformation residual calculated based on the registration anchor point to generate high-altitude adaptive registration parameters and complete sub-pixel level spatial alignment.

[0030] Step S3: Three-layer progressive fusion: The high-altitude three-layer progressive fusion module performs progressive processing: First, at the pixel level, complementary enhancement of energy distribution is performed on thermal imaging and visible light images, while high-frequency noise components in the visible light channel are dynamically suppressed according to the ultraviolet light intensity value to generate a dual-channel fused image; Second, at the feature level, thermal gradient variation features and visual semantic association features of the fused image are extracted, and a pressure correction factor is superimposed to normalize the thermal gradient amplitude to construct a three-modal joint feature vector; Finally, at the decision level, the joint feature vector is dynamically weighted, where the lower the pressure value, the more adaptively the weight of the thermal imaging channel is increased, and the higher the ultraviolet light intensity value, the more adaptively the weight of the visible light channel is decreased, generating a fusion decision map with high-altitude thermal anomaly sensitivity and visual interpretability, which is output to the multi-dimensional thermal runaway verification engine; In this embodiment of the invention, at the pixel level, the thermal imaging temperature field and the visible light image are enhanced by complementary energy. Simultaneously, high-frequency noise in the visible light channel is dynamically suppressed based on ultraviolet light intensity, filtering artifacts caused by strong ultraviolet interference at high altitudes. At the feature level, after extracting thermal gradient variation features and battery cluster visual semantic features, an air pressure correction factor is superimposed to normalize the thermal gradient amplitude, eliminating the influence of low-pressure environment on the thermal diffusion rate. At the decision level, an innovative dynamic weight allocation mechanism is established: the lower the air pressure (the higher the altitude), the higher the weight of the thermal imaging channel is automatically increased to compensate for the resolution decrease in the visible light channel due to thin air; the higher the ultraviolet light intensity, the lower the weight of the visible light channel is automatically decreased to avoid ultraviolet noise interference. Through three-layer progressive fusion, a leap from shallow image superposition to deep feature decision-making is achieved. The generated high-altitude fused decision map possesses both thermal anomaly sensitivity and visual interpretability, improving the accuracy of identifying micro-hot spots (3-5℃ temperature difference).

[0031] Step S4: Thread Verification: After receiving the fused decision graph, the multi-dimensional thermal runaway verification engine starts four verification threads in parallel: The first thread extracts the micro-hotspot connected regions in the fused decision graph and calculates the thermal diffusion compliance and temperature rise rate index of the region; the second thread performs arc flicker spectrum analysis and battery casing deformation trajectory fitting on the corresponding region in the visible light channel; the third thread calls historical thermal runaway precursor samples of the same scenario based on the energy storage power station's dedicated knowledge graph for pattern similarity comparison; the fourth thread calls the low-pressure thermal radiation propagation correction model based on the current air pressure value to perform environmental compensation calculation on the micro-hotspot temperature field; if and only if the verification results of all four threads pass the preset logic threshold and meet the spatiotemporal consistency constraints, the target is marked as a credible thermal runaway source, and a thermal runaway confidence rating is generated and output to the graded early warning and plateau situation prediction module; if any thread verification fails, the abnormal features are fed back to the high-altitude dynamic registration and compensation module to trigger registration parameter correction; In this embodiment of the invention, four independent verification threads are launched in parallel: the first thread identifies micro-hot spots through thermal gradient analysis and calculates the rate of temperature rise to determine whether it conforms to the dynamic characteristics of thermal runaway; the second thread captures arc discharge pulses through ultraviolet light spectrum analysis and fits the bulging deformation trajectory of the battery casing through visible light video; the third thread retrieves historical cases of precursors to thermal runaway based on a knowledge graph and compares the pattern similarity; the fourth thread calls a low-pressure thermal radiation correction model based on the current air pressure value to perform environmental compensation calculations on the temperature field of the micro-hot spots. Only when the results of all four threads pass the logical threshold and the spatiotemporal coordinates are consistent is it determined to be a credible thermal runaway source. If any thread fails verification, the abnormal features will be fed back to the registration module to correct the parameters, forming a verification closed loop. The four-thread collaborative verification improves false alarm suppression from a single temperature threshold judgment to a multi-dimensional cross-verification system, which can accurately identify battery thermal runaway in complex scenarios such as industrial equipment heat sources, moving reflections, and LED interference, reducing the false alarm rate.

[0032] In some embodiments, the collaborative logical relationship of the four verification threads of the multi-dimensional thermal runaway verification engine is as follows: Step S41: If the temperature rise rate index of the micro hotspot area extracted by the first thread is lower than the threshold, the target is directly identified as a regular heat source interference and the subsequent threads are terminated. Step S42: If the rate of temperature rise exceeds the threshold, the second thread is activated to perform arc flash and casing deformation analysis. When periodic arc flash or casing bulging deformation trajectory is detected in the visible light area, it is determined to be a precursor to battery thermal runaway, and the third thread is entered for verification. Step S43: The third thread performs a similarity search between the feature vector of the target area and historical thermal runaway cases in the knowledge graph of the energy storage power station. If the similarity is higher than the warning value, it is marked as a suspected thermal runaway and the confidence rating is lowered, while triggering a manual review request; if the similarity is lower than the warning value, it is confirmed as a credible thermal runaway mode and the fourth thread is activated. Step S44: The fourth thread calls the low-pressure thermal radiation propagation correction model based on the current air pressure value to perform environmental compensation calculations on the micro-hot spot temperature field. If the temperature value after compensation still exceeds the critical threshold for thermal runaway, a high confidence rating is output; otherwise, it is downgraded to a low confidence warning.

[0033] Step S5: Graded Early Warning and Plateau Situation Prediction: The graded early warning and plateau situation prediction module, based on the thermal runaway confidence rating and target spatial coordinates, combined with the thermal barrier structure constraints and cabin ventilation path data in the battery cluster layout model of the energy storage power station, calls the thermal runaway propagation simulation unit under low pressure environment to generate the fire boundary simulation results for future periods, and matches them with the response level in the plateau emergency plan library, and sends a graded early warning instruction containing target location, hazard level, and spread trend to the emergency linkage and plateau effectiveness assessment module; In this embodiment of the invention, the thermal conductivity of the thermal barrier material, structural spacing parameters, and cabin ventilation path data in the battery cluster layout model of the energy storage power station are combined. The standard thermal diffusion model is corrected by calling the low-pressure thermal convection efficiency attenuation coefficient table, generating a temperature field boundary evolution prediction that conforms to the actual conditions of the plateau environment. When the simulation results show that the fire boundary is about to touch the adjacent battery module, the warning level is automatically upgraded to the emergency response level, and personnel evacuation routes are dynamically planned, taking into account the differences in smoke deposition characteristics caused by low air pressure. Beneficial effects: This mechanism ensures that the warning level accurately matches the actual hazard level, avoiding the resource waste caused by the overestimation of fire spread speed in the normal pressure model. It also provides minute-level spread prediction, buying precious time for personnel evacuation and precise firefighting, and improving the scientific rigor and safety of emergency response.

[0034] In some embodiments, the low-pressure thermal runaway propagation simulation unit of the graded early warning and plateau situation prediction module processes the following steps: Step S51: Obtain the thermal conductivity of the thermal barrier material and the structural spacing parameters of each battery module in the battery cluster layout model; Step S52: Obtain cabin ventilation path data, including air inlet location, exhaust fan speed, and airflow direction vector; Step S53: Query the low-pressure thermal convection efficiency attenuation coefficient table based on the current air pressure value, and correct the convective heat transfer term in the standard thermal diffusion model; Step S54: Input the coordinates of the thermal runaway source, the corrected thermal diffusion model, the battery cluster layout constraints and ventilation path data into the time series simulation engine to generate the boundary evolution results of the fire temperature field for future periods; Step S55: Compare the temperature field boundary evolution results with the critical temperature of battery cluster combustion and explosion. When the predicted boundary touches the adjacent battery module, automatically upgrade the warning level to the emergency response level.

[0035] Step S6: Emergency Response and High-Altitude Performance Assessment: After receiving the graded early warning command, the emergency response and high-altitude performance assessment module performs progressive handling according to the response level: it prioritizes activating the battery cluster fire suppression device associated with the field of view for directional spraying; it simultaneously triggers audible and visual alarms and dynamic planning of evacuation routes for power station personnel; it receives feedback pressure and flow data from the fire suppression device in real time, and, combined with the energy attenuation curve of the fire source area in the fusion decision map, calls the air pressure correction factor to assess the fire suppression performance; when the temperature gradient in the fire source area is continuously decreasing and the visual smoke concentration is below the preset threshold, it is determined to be the initial fire extinguishing and enters the continuous monitoring state; if the temperature gradient is detected to rise again or the target area is split and spread, a secondary response upgrade is initiated, and the real-time handling data is transmitted back to the energy storage power station's dedicated knowledge graph for case updates; In this embodiment of the invention, after the fire suppression device is activated, the scheme collects the extinguishing agent concentration distribution, fire source energy decay curve, and nozzle pressure and flow data in real time, and determines in parallel whether the three meet the preset effectiveness threshold. The effectiveness output data is normalized by a pressure correction factor and coupled with the energy decay curve for trend analysis: if the two are positively correlated and the energy dispersion decreases, it is determined to be effective suppression; if they are negatively correlated or uncorrelated, it is determined to be suppression failure, and a secondary release of a large amount is automatically triggered. After successful fire suppression, a continuous monitoring mode is entered. If a temperature gradient rise, arc recurrence, or accelerated deformation is detected, it is determined to be reignition, and an emergency release process is immediately initiated. This mechanism achieves a leap from "open-loop release" to "closed-loop control," enabling real-time identification of fire suppression failure or pre-reignition signs, shortening the secondary response time to the second level, and avoiding the risk of "loss of control after release." Simultaneously, the effectiveness data is stored in a knowledge graph, providing data support for subsequent optimization.

[0036] In some embodiments, the barometric pressure correction factor evaluation step of the emergency response and high-altitude performance evaluation module is as follows: Step S61: After the fire suppression device is activated, the average energy value and energy distribution dispersion of the fire source area in the fusion decision map are collected at fixed time intervals to construct an energy decay time series; Step S62: Synchronously collect the real-time pressure and flow values ​​of the fire suppression device to construct an efficiency output time series; Step S63: Based on the current air pressure value, query the correction coefficient for the atomization efficiency of the extinguishing agent under low air pressure environment, and normalize the performance output time series. Step S64: Input the normalized performance output time series and energy decay time series into the trend coupling analysis unit: If the two show a positive correlation and the energy distribution dispersion continues to decrease, it is determined to be effective suppression and the current response level is maintained; if the two show a negative correlation or no correlation and the energy distribution dispersion increases, it is determined to be suppression failure, the response level is automatically upgraded and the fire suppression strategy is switched.

[0037] In some embodiments, the continuous updating mechanism of the energy storage power station-specific knowledge graph is as follows: Step S601: Package the output results of each multi-dimensional thermal runaway verification engine, the handling data of the emergency response and high-altitude performance assessment module, and the final manual confirmation label into structured case data, and label the current air pressure value and ultraviolet light intensity value as environmental context; Step S602: Perform battery module model clustering and thermal runaway mode classification on the structured case data, automatically generate new thermal runaway precursor mode nodes or low-pressure environment interference nodes, and establish causal relationship edges between nodes and correlation attributes with pressure correction coefficients. Step S603: When the number of knowledge graph nodes increases by a preset threshold, the high-altitude model parameter adaptive optimization process is triggered, and the updated graph relationship is mapped to the logical threshold parameters of the multi-dimensional thermal runaway verification engine and the air pressure compensation coefficient of the high-altitude dynamic registration and compensation module, so as to realize the closed-loop iteration of the system's discrimination capability in high-altitude scenarios.

[0038] In this embodiment of the invention, the structured five-tuple data (release command, monitoring data, judgment result, manual label, environmental context) of each early warning and fire suppression event is subjected to three-dimensional clustering based on battery type, micro-hotspot cause, and altitude range. This automatically generates new thermal runaway mode nodes or low-pressure interference nodes and establishes causal relationship edges between nodes. When the increase in the number of graph nodes reaches a threshold, an adaptive parameter optimization process is triggered, mapping the updated correction coefficients and judgment thresholds to the recognition engine, registration module, and decision model, enabling continuous iteration of system performance based on running data. This mechanism gives the system self-learning capabilities, eliminating the need for repeated manual debugging after deployment. It can adapt to different battery types and altitude scenarios, and the early warning accuracy shows a convergent upward trend after long-term operation, solving the bottleneck of performance stagnation in traditional systems and improving the average accuracy over the system's lifecycle.

[0039] Step S7: Adaptive Closed-Loop Adjustment: During system operation, the environmental parameter adaptive unit continuously monitors the rate of change of air pressure, the rate of change of ultraviolet light intensity, and the rate of change of temperature and humidity. When any rate of change exceeds the preset high-altitude stable range, the recalibration process of the high-altitude dynamic registration and compensation module is automatically triggered, realizing minute-level adaptive closed-loop adjustment of system parameters in high-altitude environments.

[0040] In this embodiment of the invention, the system presets a "high-altitude stable range" (e.g., air pressure change rate ±5 hPa / h, ultraviolet light change rate ±15% / h, temperature and humidity change rate ±3℃ / h). When the change rate of any parameter exceeds this range, the environment is determined to have entered a "non-steady state," and traditional fixed calibration parameters may become invalid. Once the over-limit judgment is triggered, the adaptive unit immediately sends a recalibration command to the high-altitude dynamic registration and compensation module. This module regenerates the compensation transformation matrix based on the latest air pressure value, reselects the registration reference mode according to the current ultraviolet intensity, and recalibrates the noise impact weight of temperature and humidity on thermal imaging. The entire process requires no manual intervention and completes the update of all system parameters within minutes (usually 1-3 minutes), ensuring that the early warning model always matches the current environmental state.

[0041] Based on the foregoing embodiments, this application provides a fire early warning system for high-altitude energy storage power stations. The various modules and units included in the system can be implemented by a processor in a computer device; of course, they can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0042] This application provides a fire early warning system for a high-altitude energy storage power station. Figure 2 A schematic diagram of a fire early warning system for a high-altitude energy storage power station provided in this application embodiment is shown below. Figure 2 As shown, this embodiment of the invention provides a fire early warning system for high-altitude energy storage power stations, comprising: The acquisition module, including a thermal imaging sensor, a visible light camera, and a barometric pressure sensor, is configured as a raw data stream acquisition device with timestamp synchronization. The high-altitude dynamic registration and compensation module includes a barometric compensation transformation unit, an ultraviolet interference determination unit, a registration reference selection unit, and a spatial alignment calculation unit, which are used to receive three-mode data and output high-altitude adaptive registration parameters. The high-altitude three-layer progressive fusion module includes a pixel-level noise suppression unit, a feature-level air pressure normalization unit, and a decision-level dynamic weight allocation unit, which are used to generate a fusion decision map. The multi-dimensional thermal runaway verification engine includes a micro hotspot analysis thread, an arc deformation analysis thread, a knowledge graph comparison thread, and a low-pressure correction thread, as well as a logic threshold unit for comprehensive judgment, which is used to output the thermal runaway confidence rating. The graded early warning and plateau situation prediction module includes a low-pressure thermal runaway propagation simulation unit and a plateau emergency plan matching unit, which are used to generate graded early warning instructions; The emergency response and high-altitude performance evaluation module includes a battery cluster fire suppression device control unit, a pressure correction performance evaluation unit, and a secondary response upgrade unit, which are used to perform progressive handling and generate feedback data. A dedicated knowledge graph library for energy storage power stations is bidirectionally connected to the multi-dimensional thermal runaway verification engine, emergency response and high-altitude performance evaluation module. It is used to store historical thermal runaway cases, low-pressure interference modes and handling data, and receive data for dynamic updates. The environmental parameter adaptive unit includes a barometric pressure sensor, an ultraviolet light sensor, a temperature and humidity sensor, and a rate of change determination subunit, which is used to monitor high-altitude environmental parameters in real time and trigger the system recalibration process. Specifically, the output of the environmental parameter adaptive unit is connected to the input of the high-altitude dynamic registration and compensation module, the output of the energy storage power station dedicated knowledge graph library is connected to the input of the multi-dimensional thermal runaway verification engine, and the feedback of the emergency linkage and plateau performance evaluation module is connected to the input of the energy storage power station dedicated knowledge graph library.

[0043] In some embodiments, the air pressure compensation transformation unit of the high-altitude dynamic registration and compensation module includes: The air pressure deviation calculation subunit is used to compare the real-time air pressure value with the standard atmospheric pressure to generate a deviation ratio; the focal length compensation coefficient query subunit is used to retrieve a preset air pressure-focal length offset mapping table based on the deviation ratio. The matrix coupling operation subunit is used to couple the focal length compensation coefficient with the spatial transformation residual to generate high-altitude adaptive registration parameters.

[0044] In some embodiments, the barometric pressure correction performance evaluation unit of the emergency response and high-altitude performance evaluation module includes: The energy decay monitoring subunit is used to extract the average energy value and distribution dispersion of the fire source area from the fusion decision map. The performance output monitoring subunit is used to collect real-time pressure and flow data of the fire suppression device; The low-pressure atomization correction subunit is used to normalize the performance output data based on the current air pressure value. The trend coupling analysis subunit is used to compare the correlation between the energy decay time series and the normalized performance output series, and output the judgment result of whether the suppression is effective or ineffective.

[0045] In some embodiments, it further includes: Power station-level edge computing nodes are deployed locally in each energy storage power station to perform real-time calculations for the high-altitude dynamic registration and compensation module, the high-altitude three-layer progressive fusion module, and the emergency response and plateau performance assessment module. The regional cloud analysis platform is connected to each power station-level edge computing node through a high-bandwidth, low-latency network. It is used to perform complex calculations and data storage for the multi-dimensional thermal runaway verification engine, the graded early warning and plateau situation prediction module, and the energy storage power station-specific knowledge graph library. The edge computing node and the cloud analysis platform exchange fusion decision graphs and thermal runaway verification results through a lightweight data transmission protocol, realizing closed-loop processing of edge-cloud collaboration in high-altitude energy storage power station scenarios.

[0046] It should be noted that, in the embodiments of this application, if the above-mentioned early warning method is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0047] Accordingly, this application provides a storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the steps in the early warning method provided in the above embodiments.

[0048] This application provides an electronic device; Figure 3 This is a schematic diagram of the composition structure of the electronic device provided in the embodiments of this application, such as... Figure 3 As shown, the electronic device 400 includes: a processor 401, at least one communication bus 402, a user interface 403, at least one external communication interface 404, and a memory 405. The communication bus 402 is configured to enable communication between these components. The user interface 403 may include a display screen, and the external communication interface 404 may include standard wired and wireless interfaces. The processor 401 is configured to execute a program of an early warning method stored in the memory to implement the steps of the early warning method provided in the above embodiment.

[0049] It should be noted that the descriptions of the storage media and electronic device embodiments above are similar to the descriptions of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage media and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0050] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0051] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, object, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, object, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, object, or apparatus that includes that element.

[0052] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0053] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0054] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0055] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0056] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a controller to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0057] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A fire early warning method for high-altitude energy storage power stations, characterized in that, The following processing steps are included: Data acquisition: The thermal imaging sensor, visible light camera and barometric pressure sensor acquire raw data streams using a preset timestamp synchronization mechanism, and input the three-modal data into the high-altitude dynamic registration and compensation module; Dynamic registration and compensation: The high-altitude dynamic registration and compensation module acquires barometric pressure sensor data in real time, calculates the offset of the thermal imager's optical parameters corresponding to the current barometric pressure value, and generates a barometric pressure compensation transformation matrix; it simultaneously acquires the temperature and humidity deviation value, illumination fluctuation value, and ultraviolet light intensity value output by the environmental parameter adaptive unit. When the ultraviolet light intensity value exceeds the preset interference threshold, the thermal imaging mode is locked as the registration reference, and the stable heat source contour in the thermal image is extracted as the registration anchor point. When the ultraviolet light intensity value is lower than the interference threshold and the barometric pressure value is in a stable range, the visible light mode is locked as the registration reference, and the structural edge features in the visible light image are extracted as the registration anchor point. Based on the selected registration anchor point and the air pressure compensation transformation matrix, the spatial transformation residual of the current frame relative to the reference template is calculated, high-altitude adaptive registration parameters are generated, and pixel-level spatial alignment of the three-modal data is completed; the registered three-modal data is then input into the high-altitude three-layer progressive fusion module. Three-layer progressive fusion: The high-altitude three-layer progressive fusion module performs progressive processing: First, at the pixel level, complementary energy distribution enhancement is performed on thermal imaging and visible light images, while high-frequency noise components in the visible light channel are dynamically suppressed according to the ultraviolet light intensity value, generating a dual-channel fused image; Second, at the feature level, thermal gradient variation features and visual semantic association features of the fused image are extracted, and a pressure correction factor is superimposed to normalize the thermal gradient amplitude, constructing a three-modal joint feature vector; Finally, at the decision level, the joint feature vector is dynamically weighted, where the lower the pressure value, the more adaptively the weight of the thermal imaging channel is increased, and the higher the ultraviolet light intensity value, the more adaptively the weight of the visible light channel is decreased, generating a fusion decision map with high-altitude thermal anomaly sensitivity and visual interpretability, which is output to the multi-dimensional thermal runaway verification engine; Thread verification: After receiving the fused decision graph, the multi-dimensional thermal runaway verification engine starts four verification threads in parallel: The first thread extracts the micro-hotspot connected regions in the fused decision graph and calculates the thermal diffusion compliance and temperature rise rate exponent of the region. The second thread performs arc scintillation spectrum analysis and battery casing deformation trajectory fitting on the corresponding region in the visible light channel. The third thread uses a knowledge graph specific to energy storage power stations to call historical thermal runaway precursor samples in the same scenario for pattern similarity comparison; the fourth thread calls a low-pressure thermal radiation propagation correction model based on the current air pressure value to perform environmental compensation calculations on the micro hotspot temperature field; if and only if the verification results of all four threads pass the preset logic threshold and meet the spatiotemporal consistency constraints, the target is marked as a credible thermal runaway source, and a thermal runaway confidence rating is generated and output to the graded early warning and plateau situation prediction module; If any thread fails to verify, the abnormal characteristics will be fed back to the high-altitude dynamic registration and compensation module, triggering the registration parameter correction. Tiered early warning and plateau situation prediction: The tiered early warning and plateau situation prediction module, based on the thermal runaway confidence rating and target spatial coordinates, combined with the thermal barrier structure constraints and cabin ventilation path data in the battery cluster layout model of the energy storage power station, calls the thermal runaway propagation simulation unit under low pressure environment to generate fire boundary projection results for future periods, and matches them with the response level in the plateau emergency plan library, and sends a tiered early warning instruction containing target location, hazard level, and spread trend to the emergency linkage and plateau effectiveness assessment module; Emergency Response and Plateau Performance Assessment: After receiving a graded early warning instruction, the emergency response and plateau performance assessment module performs progressive actions according to the response level: it prioritizes activating the battery cluster fire suppression device associated with the field of view for directional spraying; Simultaneously trigger audible and visual alarms and dynamically plan evacuation routes for power plant personnel; The system receives real-time feedback pressure and flow data from the fire suppression device, combines the energy attenuation curve of the fire source area with the fusion decision graph, and uses the air pressure correction factor to evaluate the fire suppression effectiveness. When the temperature gradient in the fire source area is continuously decreasing and the visual smoke concentration is below the preset threshold, it is determined that the initial fire has been extinguished and enters a continuous monitoring state. If the temperature gradient is detected to rise again or the target area is split and spread, a secondary response upgrade is initiated, and the real-time handling data is transmitted back to the dedicated knowledge graph of the energy storage power station for case updates. Adaptive closed-loop adjustment: During system operation, the environmental parameter adaptive unit continuously monitors the rate of change of air pressure, the rate of change of ultraviolet light intensity, and the rate of change of temperature and humidity. When any rate of change exceeds the preset high-altitude stable range, the recalibration process of the high-altitude dynamic registration and compensation module is automatically triggered, realizing minute-level adaptive closed-loop adjustment of system parameters in high-altitude environments.

2. The method according to claim 1, characterized in that, The processing steps of the high-altitude dynamic registration and compensation module specifically include: Establish initial calibration matrices for the external and internal parameters of the thermal imaging sensor, visible light camera, and barometric pressure sensor, and store them as a plateau reference registration template; While acquiring the current frame data in real time, the barometric pressure sensor data is acquired simultaneously, the deviation ratio between the current barometric pressure value and the standard atmospheric pressure is calculated, and converted into the focal length compensation coefficient of the thermal imager's optical lens. When the ultraviolet light intensity value exceeds the preset interference threshold, the thermal imaging mode is locked as the registration reference, the stable heat source contour in the thermal imaging is extracted as the registration anchor point, and the ultraviolet noise interference of the visible light channel is ignored. When the ultraviolet light intensity value is lower than the interference threshold and the air pressure deviation ratio is less than the preset fluctuation range, the visible light mode is locked as the registration reference, and the edge features of the battery cluster structure in the visible light image are extracted as the registration anchor point. By performing matrix coupling operations on the focal length compensation coefficient and the spatial transformation residual calculated based on the registration anchor point, high-altitude adaptive registration parameters are generated, and sub-pixel-level spatial alignment is achieved.

3. The method according to claim 1, characterized in that, The collaborative logic relationship of the four verification threads of the multi-dimensional thermal runaway verification engine is as follows: If the temperature rise rate index of the micro-hotspot area extracted by the first thread is lower than the threshold, the target is directly identified as a regular heat source interference and the subsequent threads are terminated. If the rate of temperature rise exceeds the threshold, the second thread is activated to analyze arc flash and casing deformation. When periodic arc flash or casing bulging deformation trajectory is detected in the visible light area, it is determined to be a precursor to battery thermal runaway, and the third thread is entered for verification. The third thread performs a similarity search between the feature vector of the target area and historical thermal runaway cases in the knowledge graph dedicated to energy storage power stations. If the similarity is higher than the warning value, it is marked as a suspected thermal runaway and the confidence rating is reduced, while triggering a manual review request. If the similarity is lower than the warning value, it is confirmed as a reliable thermal runaway mode, and the fourth thread is activated; The fourth thread calls the low-pressure thermal radiation propagation correction model based on the current air pressure value to perform environmental compensation calculations on the temperature field of the micro-hot spot. If the temperature value after compensation still exceeds the critical threshold for thermal runaway, a high confidence rating is output; otherwise, it is downgraded to a low confidence warning.

4. The method according to claim 1, characterized in that, The processing steps of the low-pressure thermal runaway propagation simulation unit in the graded early warning and plateau situation prediction module are as follows: Obtain the thermal conductivity of the thermal barrier material and the structural spacing parameters of each battery module in the battery cluster layout model; Acquire cabin ventilation path data, including air inlet location, exhaust fan speed, and airflow direction vector; Based on the current air pressure value, consult the table of low-pressure thermal convection efficiency attenuation coefficients and correct the convective heat transfer terms in the standard thermal diffusion model. The coordinates of the thermal runaway source, the modified thermal diffusion model, the battery cluster layout constraints, and the ventilation path data are input into the time series simulation engine to generate the boundary evolution results of the fire temperature field for future periods. The temperature field boundary evolution results are compared with the critical temperature of battery cluster combustion and explosion. When the predicted boundary touches the adjacent battery module, the warning level is automatically upgraded to the emergency response level.

5. The method according to claim 1, characterized in that, The air pressure correction factor evaluation steps for the emergency response and high-altitude performance evaluation module are as follows: After the fire suppression device is activated, the average energy value and energy distribution dispersion of the fire source area in the fusion decision map are collected at fixed time intervals to construct an energy decay time series. Simultaneously collect real-time pressure and flow values ​​of the fire suppression device to construct an efficiency output time series; Based on the current air pressure value, the correction coefficient for the atomization efficiency of the extinguishing agent under low air pressure is queried, and the time series of performance output is normalized. The normalized performance output time series and energy decay time series are input into the trend coupling analysis unit: if the two show a positive correlation and the energy distribution dispersion continues to decrease, it is determined to be effective suppression and the current response level is maintained. If the two show a negative correlation or no correlation and the energy distribution dispersion increases, it is determined to be a suppression failure, and the response level is automatically upgraded and the fire suppression strategy is switched.

6. The method according to claim 1, characterized in that, The continuous updating mechanism for the dedicated knowledge graph of the energy storage power station is as follows: The output results of each multi-dimensional thermal runaway verification engine, the handling data of the emergency response and high-altitude performance assessment module, and the final manual confirmation label are packaged into structured case data, and the current air pressure value and ultraviolet light intensity value are labeled as environmental context. The system performs battery module model clustering and thermal runaway mode classification on structured case data, automatically generating new thermal runaway precursor mode nodes or low-pressure environment interference nodes, and establishing causal relationship edges between nodes and related attributes of air pressure correction coefficients. When the number of knowledge graph nodes increases by a preset threshold, the system triggers the high-altitude model parameter adaptive optimization process, mapping the updated graph relationships to the logical threshold parameters of the multi-dimensional thermal runaway verification engine and the air pressure compensation coefficient of the high-altitude dynamic registration and compensation module, thereby realizing the closed-loop iteration of the system's discrimination capability in high-altitude scenarios.

7. A fire early warning system for a high-altitude energy storage power station, characterized in that, include: The acquisition module, including a thermal imaging sensor, a visible light camera, and a barometric pressure sensor, is configured as a raw data stream acquisition device with timestamp synchronization. The high-altitude dynamic registration and compensation module includes a barometric compensation transformation unit, an ultraviolet interference determination unit, a registration reference selection unit, and a spatial alignment calculation unit, which are used to receive three-mode data and output high-altitude adaptive registration parameters. The high-altitude three-layer progressive fusion module includes a pixel-level noise suppression unit, a feature-level air pressure normalization unit, and a decision-level dynamic weight allocation unit, which are used to generate a fusion decision map. The multi-dimensional thermal runaway verification engine includes a micro hotspot analysis thread, an arc deformation analysis thread, a knowledge graph comparison thread, and a low-pressure correction thread, as well as a logic threshold unit for comprehensive judgment, which is used to output the thermal runaway confidence rating. The graded early warning and plateau situation prediction module includes a low-pressure thermal runaway propagation simulation unit and a plateau emergency plan matching unit, which are used to generate graded early warning instructions; The emergency response and high-altitude performance evaluation module includes a battery cluster fire suppression device control unit, a pressure correction performance evaluation unit, and a secondary response upgrade unit, which are used to perform progressive handling and generate feedback data. A dedicated knowledge graph library for energy storage power stations is bidirectionally connected to the multi-dimensional thermal runaway verification engine, emergency response and high-altitude performance evaluation module. It is used to store historical thermal runaway cases, low-pressure interference modes and handling data, and receive data for dynamic updates. The environmental parameter adaptive unit includes a barometric pressure sensor, an ultraviolet light sensor, a temperature and humidity sensor, and a rate of change determination subunit, which is used to monitor high-altitude environmental parameters in real time and trigger the system recalibration process. Specifically, the output of the environmental parameter adaptive unit is connected to the input of the high-altitude dynamic registration and compensation module, the output of the energy storage power station dedicated knowledge graph library is connected to the input of the multi-dimensional thermal runaway verification engine, and the feedback of the emergency linkage and plateau performance evaluation module is connected to the input of the energy storage power station dedicated knowledge graph library.

8. The system according to claim 7, characterized in that, The air pressure compensation transformation unit of the high-altitude dynamic registration and compensation module includes: The air pressure deviation calculation subunit is used to compare the real-time air pressure value with the standard atmospheric pressure to generate a deviation ratio; the focal length compensation coefficient query subunit is used to retrieve a preset air pressure-focal length offset mapping table based on the deviation ratio. The matrix coupling operation subunit is used to couple the focal length compensation coefficient with the spatial transformation residual to generate high-altitude adaptive registration parameters.

9. The system according to claim 7, characterized in that, The air pressure correction performance evaluation unit of the emergency response and high-altitude performance evaluation module includes: The energy decay monitoring subunit is used to extract the average energy value and distribution dispersion of the fire source area from the fusion decision map. The performance output monitoring subunit is used to collect real-time pressure and flow data of the fire suppression device; The low-pressure atomization correction subunit is used to normalize the performance output data based on the current air pressure value. The trend coupling analysis subunit is used to compare the correlation between the energy decay time series and the normalized performance output series, and output the judgment result of whether the suppression is effective or ineffective.

10. The system according to claim 7, characterized in that, Further includes: Power station-level edge computing nodes are deployed locally in each energy storage power station to perform real-time calculations for the high-altitude dynamic registration and compensation module, the high-altitude three-layer progressive fusion module, and the emergency response and plateau performance assessment module. The regional cloud analysis platform is connected to each power station-level edge computing node through a high-bandwidth, low-latency network. It is used to perform complex calculations and data storage for the multi-dimensional thermal runaway verification engine, the graded early warning and plateau situation prediction module, and the energy storage power station-specific knowledge graph library. The edge computing node and the cloud analysis platform exchange fusion decision graphs and thermal runaway verification results through a lightweight data transmission protocol, realizing closed-loop processing of edge-cloud collaboration in high-altitude energy storage power station scenarios.

Images