Energy storage power station fire early warning method and system based on multi-parameter fusion
The energy storage power station fire warning method based on multi-parameter fusion and dynamic risk grading solves the problem of insufficient early feature recognition of lithium battery thermal runaway in existing technologies, achieves accurate warning and timely response, and reduces the risk of fire accidents.
Patent Information
- Application Number
- CN202510998473.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-18
- Publication Date
- 2025-09-26
AI Technical Summary
Existing fire warning technology for energy storage power stations relies on single parameter monitoring or simple logic superposition, which cannot accurately capture the early characteristics of lithium battery thermal runaway, resulting in false alarms or delayed warnings, making it difficult to prevent fire accidents.
Using a multi-parameter fusion method, distributed sensors are used to collect real-time data on lithium battery temperature, characteristic gas concentration, battery cell expansion force, and ambient temperature and humidity. Combined with the environmental compensation algorithm and random forest algorithm, deep fusion analysis is performed to build a four-level dynamic risk grading mechanism to trigger precise early warnings and linked emergency responses.
It has achieved accurate identification of fire hazards under complex working conditions, reduced false alarm rates, improved the timeliness and practicality of early warnings, ensured that targeted intervention can be triggered at the early stage of thermal runaway, and minimized the losses caused by fire accidents.
Smart Images

Figure CN120708344A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of fire safety technology, and in particular to a fire early warning method and system for an energy storage power station based on multi-parameter fusion. Background Art
[0002] The fire warning system for energy storage power stations is a system that detects fire hazards in advance by comprehensively monitoring battery status and external environmental parameters. It mainly targets the thermal runaway risks and electrical fire hazards of lithium battery energy storage systems. The core technologies of the online monitoring and fire warning system for energy storage power stations mainly include data acquisition and processing, data analysis and diagnosis, fire warning and emergency response. First, data acquisition and processing are the foundation of the system. By real-time collection and transmission of key information such as the internal environment, equipment status, and operating parameters of the energy storage power station, the accuracy and timeliness of the data are ensured. Secondly, data analysis and diagnosis are the core of the system. By in-depth mining and analysis of the collected data, potential safety hazards and failure modes are discovered, providing strong support for fire warning. Finally, fire warning and emergency response are the goals of the system. Once a fire risk or abnormal situation is detected, the system will immediately issue a warning signal and initiate corresponding emergency response measures to minimize the losses caused by fire accidents.
[0003] Existing fire warning technologies for energy storage power stations mostly rely on single-parameter monitoring or simply perform simple logical superposition of multiple parameters, without combining environmental factor compensation and intelligent algorithms to achieve deep dynamic fusion. These technologies are unable to accurately capture correlated features such as sudden changes in temperature gradients, trace release of characteristic gases, and increasing cell expansion forces in the early stages of lithium battery thermal runaway. Therefore, they are susceptible to false alarms due to environmental interference under complex working conditions, or are delayed in warnings due to failure to identify early weak signals, resulting in untimely fire prevention and control responses and difficulty in blocking the thermal runaway chain reaction, ultimately triggering fire accidents in energy storage power stations, causing equipment damage, economic losses, and even casualties. Summary of the Invention
[0004] In view of the shortcomings of the existing technology, the present invention provides a fire warning method and system for an energy storage power station based on multi-parameter fusion, which solves the problems mentioned in the background technology.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a fire early warning method for energy storage power stations based on multi-parameter fusion, comprising the following steps:
[0006] A1. Use distributed sensors to collect real-time data on the temperature, characteristic gas concentration, cell expansion force, voltage fluctuation, and ambient temperature and humidity of lithium batteries in energy storage power stations.
[0007] A2. Clean, denoise, and standardize the collected parameters. Use the ambient temperature and humidity compensation algorithm to correct the temperature and gas concentration monitoring values and eliminate abnormal data caused by environmental interference.
[0008] A3. Input the pre-processed data into a preset intelligent warning module. The model is based on a random forest algorithm and integrates the temperature gradient change rate, characteristic gas concentration threshold, expansion force mutation coefficient, voltage consistency deviation, and environmental impact weight for comprehensive calculation.
[0009] A4. Based on the model output results, fire risks are divided into four levels: normal, warning, alarm, and emergency. Each level corresponds to a dynamically adjusted parameter threshold. When the temperature gradient is greater than 5°C / min, the CO concentration is greater than 50ppm, the expansion force mutation is greater than 10N, or the voltage consistency deviation is greater than 5%, an early warning is triggered. When the parameters exceed 2 times the above threshold, an emergency warning is triggered and the highest level of emergency response is linked.
[0010] Preferably, the training process of the intelligent early warning module includes:
[0011] Historical normal operation data and thermal runaway simulation experiment data of different battery types are collected to construct a multi-type sample set. An improved random forest algorithm is adopted, and an attention mechanism is introduced to assign dynamic weights to key parameters. The model parameters are optimized through 5-fold cross-validation to improve the accuracy of early warning and reduce the false alarm rate.
[0012] Preferably, the ambient temperature and humidity compensation algorithm in step A2 is specifically as follows:
[0013] A correction formula is constructed based on the ambient temperature and relative humidity, and the original temperature monitoring value is corrected to:
[0014] T'=T0×(1+0.002×(25-T))×(1+0.001×(50-RH))
[0015] Where RH is the relative humidity of the environment, T is the ambient temperature, and T0 is the original temperature monitoring value;
[0016] The gas concentration monitoring value is corrected to:
[0017] C'=C0×(1+0.003×(RH-50))
[0018] Among them, C0 is the gas concentration monitoring value. The above correction calculation of temperature and gas concentration is used to eliminate the influence of environmental factors on sensor accuracy.
[0019] Preferably, the thermal runaway propagation path prediction step: based on the multi-parameter fusion analysis results, combined with the battery module arrangement structure and heat conduction model, predicts the possible propagation path and time of thermal runaway, and synchronously outputs the propagation range prediction map in the early warning signal to provide a basis for directional fire extinguishing.
[0020] A fire warning system for an energy storage power station based on multi-parameter fusion, the fire warning system for an energy storage power station comprising:
[0021] Distributed monitoring module: It consists of sensors placed in the battery cells, modules, cabins and stations, and is used to collect temperature, characteristic gas concentration, battery cell expansion force, voltage fluctuation and ambient temperature and humidity data;
[0022] Data fusion processing module: This module uses a collaborative architecture between edge computing nodes and the cloud. The edge nodes are responsible for real-time data preprocessing and rapid early warning, while the cloud is responsible for model training and updating and global data analysis. The two exchange data through encrypted wireless communication.
[0023] Intelligent early warning module: supports online model updates and can dynamically adjust parameter thresholds based on battery cycle times;
[0024] Linkage response module: Receives signals from the intelligent early warning module, triggers graded sound and light alarms, links the fire protection system and the power grid dispatching system, automatically sends a load reduction request to the power grid in an emergency, and visualizes the early warning and disposal process through the digital twin platform.
[0025] Preferably, the sensor installed on the battery cell integrates a micro pressure and voltage composite probe, and the cabin is equipped with an infrared thermal imager to achieve real-time imaging of the regional temperature field;
[0026] The sensors of the distributed monitoring module are redundantly set, and the sensor redundancy protection is set as follows: dual primary and backup sensors are configured at key monitoring points. When the primary sensor fails or the data deviation is greater than the set threshold, it automatically switches to the backup sensor to ensure data collection continuity.
[0027] Preferably, the intelligent early warning module includes: a battery health correlation analysis submodule;
[0028] The battery health correlation analysis submodule links the fire warning model with the battery SOH assessment model. When the battery SOH is less than 80%, the warning sensitivity is automatically increased, and abnormal parameters of aging batteries are monitored first.
[0029] Preferably, the fire linkage mode of the linkage response module is: based on the prediction result of the thermal runaway propagation path, the fire extinguishing device within a range of 3 meters around the fire point is activated in a targeted manner to avoid secondary damage to the equipment caused by the full-area fire extinguishing;
[0030] At the same time, parameter changes during the fire extinguishing process are recorded and fed back to the intelligent early warning module for model optimization.
[0031] Preferably, the energy storage power station fire warning system further includes: a multi-cabin collaborative warning module;
[0032] The multi-cabin collaborative early warning module automatically sends an early warning signal to the adjacent cabin when the corresponding prefabricated cabin triggers an alarm. The monitoring modules of the adjacent cabins increase the sampling frequency and activate the fire isolation doors between the cabins to block the fire spread channel.
[0033] Preferably, the data fusion processing module supports data interaction with the power management system of the energy storage power station, integrates the operating data related to the charge and discharge power and the number of cycles, and automatically enhances the monitoring weight of the voltage fluctuation and temperature gradient when the charge and discharge rate is greater than 1C, so as to adapt to the early warning requirements under high load conditions.
[0034] The present invention provides a fire warning method and system for energy storage power plants based on multi-parameter fusion. It has the following beneficial effects:
[0035] (1) By collecting multi-dimensional parameters such as temperature, characteristic gas concentration, cell expansion force, voltage fluctuation and ambient temperature and humidity, and combining the ambient temperature and humidity compensation algorithm to correct the monitoring value to eliminate environmental interference, the random forest algorithm is used to perform a deep fusion analysis of the correlation characteristics of each parameter, rather than relying on a single parameter or simple logical superposition. It can accurately capture the temperature gradient mutation, characteristic gas trace release and other related signals in the early stage of thermal runaway of lithium batteries, avoid the interference of environmental factors on monitoring accuracy, and effectively solve the false alarm problem caused by single parameters or insufficient fusion in the existing technology, ensuring that fire hazards can be accurately identified even under complex working conditions, providing a reliable basis for early intervention, and improving the overall reliability of fire warning of energy storage power stations.
[0036] (2) Based on the results of multi-parameter fusion, a four-level dynamic risk grading mechanism is constructed. Each level corresponds to a dynamically adjusted parameter threshold, and dynamic weights are assigned to key parameters through the intelligent early warning model. It can accurately match the warning level according to the different stages of thermal runaway development, avoiding the warning lag caused by fixed thresholds and adapting to the disposal needs of different risk levels through graded responses. It realizes accurate early warning from the early stage of thermal runaway to the emergency state, ensuring that the early warning can be triggered at the risk budding stage, buying time to block the thermal runaway chain reaction, and reducing the interference of excessive early warnings on the normal operation of the power station, thereby improving the timeliness and practicality of the early warning.
[0037] (3) The multi-parameter fusion analysis is linked with the emergency response, which can trigger the highest level of response in an emergency, and support targeted disposal in combination with the parameter change characteristics, forming a closed loop from monitoring, analysis, early warning to response, rather than just staying at a single early warning level. It can quickly convert early warning information into specific prevention and control actions, avoiding response delays caused by the disconnection between early warning and disposal, and effectively opening up the entire process of fire prevention and control, ensuring that targeted intervention can be initiated before thermal runaway spreads, minimizing damage to equipment caused by fire accidents, reducing economic losses, and providing full-chain protection for the safe operation of energy storage power stations. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 This is a method flow chart of a fire warning method and system for an energy storage power station based on multi-parameter fusion according to the present invention;
[0039] Figure 2 This is a system block diagram of a fire warning method and system for an energy storage power station based on multi-parameter fusion according to the present invention. DETAILED DESCRIPTION
[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0041] Example 1
[0042] See also Figure 1-2 The present invention provides a fire warning method and system for energy storage power stations based on multi-parameter fusion. To achieve the above objectives, the present invention is implemented through the following technical solutions, including the following steps:
[0043] A1. Use distributed sensors to collect real-time data on the temperature, characteristic gas concentration, cell expansion force, voltage fluctuation, and ambient temperature and humidity of lithium batteries in energy storage power stations.
[0044] A2. Clean, denoise, and standardize the collected parameters. Use the ambient temperature and humidity compensation algorithm to correct the temperature and gas concentration monitoring values and eliminate abnormal data caused by environmental interference.
[0045] A3. Input the pre-processed data into the preset intelligent warning module. The model is based on the random forest algorithm and integrates the temperature gradient change rate, characteristic gas concentration threshold, expansion force mutation coefficient, voltage consistency deviation, and environmental impact weight for comprehensive calculation.
[0046] A4. Based on the model output results, fire risks are divided into four levels: normal, warning, alarm, and emergency. Each level corresponds to a dynamically adjusted parameter threshold. When the temperature gradient is greater than 5°C / min, the CO concentration is greater than 50ppm, the expansion force mutation is greater than 10N, or the voltage consistency deviation is greater than 5%, an early warning is triggered. When the parameters exceed 2 times the above threshold, an emergency warning is triggered and the highest level of emergency response is linked.
[0047] A fire warning system for an energy storage power station based on multi-parameter fusion, comprising:
[0048] Distributed monitoring module: It consists of sensors placed in the battery cells, modules, cabins and stations, and is used to collect temperature, characteristic gas concentration, battery cell expansion force, voltage fluctuation and ambient temperature and humidity data;
[0049] Data fusion processing module: This module uses a collaborative architecture between edge computing nodes and the cloud. The edge nodes are responsible for real-time data preprocessing and rapid early warning, while the cloud is responsible for model training and updating and global data analysis. The two exchange data through encrypted wireless communication.
[0050] Intelligent early warning module: supports online model updates and can dynamically adjust parameter thresholds based on battery cycle times;
[0051] Linkage response module: Receives signals from the intelligent early warning module, triggers graded sound and light alarms, links the fire protection system and the power grid dispatching system, automatically sends a load reduction request to the power grid in an emergency, and visualizes the early warning and disposal process through the digital twin platform.
[0052] In this embodiment, the distributed monitoring module is arranged through four levels of sensors: cell level, module level, cabin level, and station level:
[0053] Cell level: A distributed optical fiber temperature sensor is attached to the surface of each battery cell, and a miniature piezoresistive expansion force sensor is embedded between the cells, connected to the module controller via a flexible circuit board;
[0054] Module level: Each battery module is equipped with a voltage patrol meter, and an optical fiber temperature measurement point is arranged at the center of the module to calculate the temperature gradient;
[0055] Cabin level: Each prefabricated cabin is equipped with a multi-gas analyzer, which is arranged at the top, middle and bottom of the cabin. It can simultaneously detect CO, H2, hydrogen fluoride and VOCs. Two temperature and humidity sensors are installed in the corners of the cabin. The multi-gas analyzers at the top, middle and bottom of the cabin are used to monitor hot smoke gas, the middle one is used to monitor lateral diffusion gas, and the bottom one is used to monitor sinking gas.
[0056] In-station level: A data aggregation gateway is deployed in the station control room to receive data from each cabin via industrial Ethernet, with a unified sampling period of 1 second.
[0057] In actual operation, the sensor collects and uploads data in real time. When a battery cell experiences an internal short circuit, the fiber optic sensor detects a rise in surface temperature from 35°C at a rate of 2°C / min, an increase in expansion force from an initial 2N to 5N, and a drop in voltage from 3.2V to 3.15V. Simultaneously, the gas sensor at the bottom of the cabin detects a slow rise in CO concentration from 0ppm to 10ppm, followed by a warning.
[0058] The edge computing unit of the data fusion processing module processes the collected data:
[0059] Outliers are eliminated using the 3σ criterion. When the temperature of a battery cell jumps by more than 20°C within 1 second, it is determined to be sensor interference. The data is automatically eliminated and the mean value of the previous 3 seconds is used for replacement. Normalization processing is performed, and all parameters are mapped to the range of 0-1 to ensure the uniformity of the model input dimension. The preprocessed data is uploaded to the intelligent early warning module, which has a built-in early warning model based on the random forest algorithm. The model input features include: temperature gradient change rate, CO concentration growth rate, expansion force mutation coefficient, voltage consistency deviation and environmental impact weight.
[0060] The model votes on features through a decision tree and outputs a comprehensive risk value. For example, when the temperature gradient of a battery cell reaches 6°C / min (corresponding to a risk value of 20), the CO concentration growth rate is 5ppm / min (risk value 15), the expansion force mutation coefficient is 1.8 (risk value 25), the voltage deviation is 4% (risk value 10), and the environmental weight is 1.1, the comprehensive risk value is (20+15+25+10)*1.1=77, and an early warning is immediately triggered.
[0061] Divide into four levels according to the comprehensive risk value:
[0062] Normal (risk value 0-30): All parameters are within the safe range and the system has no warning;
[0063] Warning (risk value 30-50): Triggers a yellow sound and light alarm, and pushes a warning message to the operation and maintenance app, indicating an abnormal rise in battery cell temperature and recommending checking the corresponding module;
[0064] Alarm (risk value 50-80): A red sound and light alarm is triggered, the fault location is highlighted on the large screen in the station control room, and the operation and maintenance personnel receive a text message notification including the location information of the abnormal area;
[0065] Emergency (risk value > 80): Automatically initiate the highest level response, cut off the main circuit breaker of the faulty cabin, link the fire protection system to start the gas fire extinguishing device, and send a load reduction request to the power grid dispatching system.
[0066] Through multi-dimensional parameter collection and deep integration, the limitations of traditional single-parameter early warning are overcome: because it covers full-level monitoring from battery cells to station levels, combined with environmental compensation algorithms, it eliminates the interference of high temperature and high humidity on the data, ensuring that early signals of thermal runaway can still be accurately captured under complex working conditions. Through the four-level dynamic graded early warning, it not only avoids the interference of excessive warnings on power station operations, but also can respond quickly when risks escalate. Compared with traditional single-parameter temperature warnings, hidden dangers are discovered 15 to 20 minutes earlier, effectively reducing the probability of fire accidents. At the same time, the various modules of the system work together to achieve closed-loop management from data collection to emergency response, providing full-process protection for the safe operation of energy storage power stations.
[0067] Example 2
[0068] Specifically: The training process of the intelligent early warning module includes: collecting historical normal operation data and thermal runaway simulation experiment data of different battery types to construct a multi-type sample set; using an improved random forest algorithm, introducing an attention mechanism to assign dynamic weights to key parameters, and optimizing model parameters through 5-fold cross-validation to improve the accuracy of early warnings and reduce false alarm rates.
[0069] The ambient temperature and humidity compensation algorithm in step A2 is specifically as follows:
[0070] A correction formula is constructed based on the ambient temperature and relative humidity, and the original temperature monitoring value is corrected to:
[0071] T'=T0×(1+0.002×(25-T))×(1+0.001×(50-RH))
[0072] Where RH is the relative humidity of the environment, T is the ambient temperature, and T0 is the original temperature monitoring value;
[0073] The gas concentration monitoring value is corrected to:
[0074] C'=C0×(1+0.003×(RH-50))
[0075] Among them, C0 is the gas concentration monitoring value. The above correction calculation of temperature and gas concentration is used to eliminate the influence of environmental factors on sensor accuracy.
[0076] Thermal runaway propagation path prediction step: Based on the results of multi-parameter fusion analysis, combined with the battery module layout structure and heat conduction model, the possible path and time of thermal runaway propagation are predicted, and a propagation range prediction map is output simultaneously with the early warning signal to provide a basis for targeted fire fighting;
[0077] In this embodiment, two types of data are collected to construct a sample set: normal operation data, covering different battery types, charge and discharge rates, environmental conditions and aging levels. Each set of data contains a multi-parameter time series curve for one consecutive hour. Thermal runaway simulation data, through laboratory simulation of battery failures such as overcharge, short circuit, and needle puncture, collects the parameters of the battery from normal to thermal runaway.
[0078] Model improvement and training: An improved random forest algorithm is used, and an attention mechanism is introduced to dynamically assign weights to key parameters. Parameters sensitive to thermal runaway are given an initial weight of 1.5, and minor parameters are given a weight of 0.8. During training, the weights are optimized using the gradient descent method. When the frequency of a parameter in thermal runaway samples is greater than 80%, the weight is automatically increased by 0.1. Conversely, if the parameter fluctuates greatly in normal samples, the weight is reduced by 0.1. Subsequently, a 5-fold cross-validation is used to randomly divide the sample set into 5 parts, 4 of which are used as training sets and 1 as a validation set. The average accuracy is calculated after 5 cycles. The final model accuracy reached 99.3%, with a false alarm rate of 0.2% and a false negative rate of 0.1%.
[0079] Ternary lithium batteries are more sensitive to ambient temperature. The environmental compensation algorithm can significantly improve monitoring accuracy. In low temperature and low humidity scenarios, such as in winter when the cabin temperature is 5°C and the humidity is 30%, the original temperature monitoring value of a battery cell is 10°C. The following formula is used to correct it:
[0080] T'=T0×(1+0.002×(25-T))×(1+0.001×(50-RH))
[0081] Make the real-time temperature of the battery cell closer to the actual temperature;
[0082] Similarly, the ambient gas concentration monitoring value is corrected to:
[0083] C'=C0×(1+0.003×(RH-50))
[0084] The interference of humidity on the gas sensor is eliminated. In actual tests, the deviation between the compensated data and the laboratory standard value decreases, providing reliable input for model analysis.
[0085] For the prediction of thermal runaway propagation path, based on the multi-parameter fusion results combined with the battery module layout and heat conduction model, the prediction steps are as follows:
[0086] Input parameters: current thermal runaway cell position, real-time temperature, temperature gradient, and initial temperature of surrounding cells;
[0087] Model calculation: Using finite element analysis software to simulate heat transfer: After 1 minute, the temperature of adjacent cells rises. After 3 minutes, the temperature of 50% of the cells in the same module rises synchronously. After 5 minutes, the heat diffuses to the adjacent modules.
[0088] Output: A colored spread path is generated, with red indicating out-of-control areas and yellow indicating high-risk areas. The path is labeled "spread to the entire area of module 3 within 5 minutes, and to modules 2 and 4 of cabin 2 within 10 minutes." This path is then sent to the fire linkage system, which then uses firefighting equipment to prevent or extinguish the fire.
[0089] Through the improved random forest model and attention mechanism, the sensitive parameters of thermal runaway of ternary lithium batteries are given higher weights, which significantly improves the accuracy of early warning and reduces the false alarm rate compared with traditional models. The environmental temperature and humidity compensation algorithm corrects the sensor deviation under different working conditions in a targeted manner, improves data accuracy, and provides a reliable basis for model analysis. The thermal runaway spread path prediction, combined with the module structure and heat conduction characteristics, can accurately predict the spread of fire, providing a scientific basis for targeted fire extinguishing, reducing the secondary damage rate of equipment compared with full-area fire extinguishing, and shortening the fire extinguishing response time, ensuring that the spread is blocked before the thermal runaway chain reaction is formed.
[0090] Example 3
[0091] Specifically: The sensor installed on the battery cell integrates a micro pressure and voltage composite probe, and an infrared thermal imager is installed in the cabin to achieve real-time imaging of the regional temperature field;
[0092] The sensors of the distributed monitoring module are redundantly set, and the sensor redundancy protection is set as follows: dual primary and backup sensors are configured at key monitoring points. When the primary sensor fails or the data deviation is greater than the set threshold, it automatically switches to the backup sensor to ensure data collection continuity.
[0093] The intelligent early warning module includes: battery health correlation analysis submodule;
[0094] Battery health correlation analysis submodule: Links the fire warning model with the battery SOH assessment model. When the battery SOH is less than 80%, the warning sensitivity is automatically increased, and abnormal parameters of aging batteries are monitored first.
[0095] The firefighting linkage mode of the linkage response module is as follows: based on the prediction results of the thermal runaway spread path, the fire extinguishing equipment within a 3-meter radius around the fire point is activated in a targeted manner to avoid secondary damage to equipment caused by full-area fire extinguishing;
[0096] At the same time, parameter changes during the fire extinguishing process are recorded and fed back to the intelligent early warning module for model optimization.
[0097] The energy storage power station fire warning system also includes: a multi-cabin collaborative warning module;
[0098] The multi-cabin collaborative early warning module automatically sends an early warning signal to the adjacent cabin when the corresponding prefabricated cabin triggers an alarm. The monitoring modules of the adjacent cabins increase the sampling frequency and activate the fire isolation doors between the cabins to block the fire spread channel.
[0099] The data fusion processing module supports data interaction with the energy storage station's power management system, integrating operational data related to charge and discharge power and cycle count. When the charge and discharge rate is greater than 1C, it automatically increases the monitoring weight of voltage fluctuations and temperature gradients to meet early warning requirements under high-load conditions.
[0100] In this embodiment, the linkage response module initiates fire extinguishing according to the following logic based on the prediction results of the thermal runaway propagation path: each prefabricated cabin is divided into three fire extinguishing zones according to the location of the battery cluster: the zones are upper, middle, and lower. Each zone is equipped with two sets of perfluorohexanone fire extinguishing devices. When the thermal runaway is predicted to spread to the left zone, only the left fire extinguishing device is activated and the spraying is completed within 30 seconds to prevent the normal modules on the right from being affected by the fire extinguishing agent. The temperature and gas concentration changes during the fire extinguishing process are recorded and fed back to the intelligent early warning module for optimization of model parameters.
[0101] The multi-cabin collaborative warning module enables inter-cabin communication via an industrial bus. The response process is as follows: When a cabin triggers an alarm-level warning with a risk value of 60, a collaborative signal is immediately sent to adjacent cabins. The collaborative signal contains the fault location and risk level. The sampling frequency of the adjacent cabin's monitoring module is increased from 1Hz to 5Hz, and the fire isolation door between cabins is closed within 10 seconds. At the same time, the warning threshold is temporarily lowered. When the risk level of the corresponding risk cabin drops to normal, a release signal is sent, and the adjacent cabin resumes normal operation.
[0102] By adopting a sensor redundancy design and utilizing a master-slave switching mechanism, the continuity of data collection is ensured, the data interruption rate is reduced compared to a single-sensor solution, and smooth and safe monitoring is guaranteed. Battery health correlation analysis adjusts the early warning strategy for aging cells in a targeted manner, thereby improving the fault detection rate of aging batteries. Aging batteries can be monitored more accurately and efficiently, ensuring the safe use of aging batteries. The directional fire extinguishing method of the fire linkage avoids full-area spraying, reduces the secondary damage rate of equipment, and reduces the consumption of fire extinguishing agents. The multi-compartment collaborative early warning quickly isolates the fault area and blocks the spread of fire across compartments, reducing the risk of accident expansion compared to independent compartment designs. The interaction with the energy management system improves the accuracy of early warnings during high-rate charging and discharging to meet the monitoring needs under high-load conditions, comprehensively improving the safety and control capabilities of the energy storage power station under complex working conditions.
[0103] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.
Claims
1. A fire warning method for energy storage power station based on multi-parameter fusion, characterized in that: The following steps are involved: A1. Use distributed sensors to collect real-time data on the temperature, characteristic gas concentration, cell expansion force, voltage fluctuation, and ambient temperature and humidity of lithium batteries in energy storage power stations. A2. Clean, denoise, and standardize the collected parameters. Use the ambient temperature and humidity compensation algorithm to correct the temperature and gas concentration monitoring values and eliminate abnormal data caused by environmental interference. A3. Input the pre-processed data into a preset intelligent warning module. The model is based on a random forest algorithm and integrates the temperature gradient change rate, characteristic gas concentration threshold, expansion force mutation coefficient, voltage consistency deviation, and environmental impact weight for comprehensive calculation. A4. Based on the model output results, fire risks are divided into four levels: normal, warning, alarm, and emergency. Each level corresponds to a dynamically adjusted parameter threshold. When the temperature gradient is greater than 5°C / min, the CO concentration is greater than 50ppm, the expansion force mutation is greater than 10N, or the voltage consistency deviation is greater than 5%, an early warning is triggered. When the parameters exceed 2 times the above threshold, an emergency warning is triggered and the highest level of emergency response is linked.
2. The fire warning method for energy storage power station based on multi-parameter fusion according to claim 1 is characterized in that: The training process of the intelligent early warning module includes: Historical normal operation data and thermal runaway simulation experiment data of different battery types are collected to construct a multi-type sample set. An improved random forest algorithm is adopted, and an attention mechanism is introduced to assign dynamic weights to key parameters. The model parameters are optimized through 5-fold cross-validation to improve the accuracy of early warning and reduce the false alarm rate.
3. The fire warning method for energy storage power station based on multi-parameter fusion according to claim 1 is characterized in that: The ambient temperature and humidity compensation algorithm in step A2 is specifically as follows: A correction formula is constructed based on the ambient temperature and relative humidity, and the original temperature monitoring value is corrected to: T'=T0×(1+0.002×(25-T))×(1+0.001×(50-RH)) Where RH is the relative humidity of the environment, T is the ambient temperature, and T0 is the original temperature monitoring value; The gas concentration monitoring value is corrected to: C ’ =C0×(1+0.003×(RH-50)) Among them, C0 is the gas concentration monitoring value. The above correction calculation of temperature and gas concentration is used to eliminate the influence of environmental factors on sensor accuracy.
4. The fire warning method for energy storage power station based on multi-parameter fusion according to claim 1 is characterized in that: The thermal runaway propagation path prediction step: Based on the multi-parameter fusion analysis results, combined with the battery module arrangement structure and heat conduction model, the possible propagation path and time of thermal runaway are predicted, and a propagation range prediction map is output simultaneously in the early warning signal to provide a basis for directional fire extinguishing.
5. A fire warning system for an energy storage power station based on multi-parameter fusion, applied to a fire warning method for an energy storage power station based on multi-parameter fusion as claimed in any one of claims 1 to 4, characterized in that: The energy storage power station fire warning system includes: Distributed monitoring module: It consists of sensors placed in the battery cells, modules, cabins and stations, and is used to collect temperature, characteristic gas concentration, battery cell expansion force, voltage fluctuation and ambient temperature and humidity data; Data fusion processing module: This module uses a collaborative architecture between edge computing nodes and the cloud. The edge nodes are responsible for real-time data preprocessing and rapid early warning, while the cloud is responsible for model training and updating and global data analysis. The two exchange data through encrypted wireless communication. Intelligent early warning module: supports online model updates and can dynamically adjust parameter thresholds based on battery cycle times; Linkage response module: Receives signals from the intelligent early warning module, triggers graded sound and light alarms, links the fire protection system and the power grid dispatching system, automatically sends a load reduction request to the power grid in an emergency, and visualizes the early warning and disposal process through the digital twin platform.
6. The energy storage power station fire warning system based on multi-parameter fusion according to claim 5 is characterized in that: The sensor installed on the battery cell integrates a micro pressure and voltage composite probe, and the cabin is equipped with an infrared thermal imager to achieve real-time imaging of the regional temperature field; The sensors of the distributed monitoring module are redundantly set, and the sensor redundancy protection is set as follows: dual primary and backup sensors are configured at key monitoring points. When the primary sensor fails or the data deviation is greater than the set threshold, it automatically switches to the backup sensor to ensure data collection continuity.
7. The energy storage power station fire warning system based on multi-parameter fusion according to claim 5 is characterized in that: The intelligent early warning module includes: a battery health correlation analysis submodule; The battery health correlation analysis submodule links the fire warning model with the battery SOH assessment model. When the battery SOH is less than 80%, the warning sensitivity is automatically increased, and abnormal parameters of aging batteries are monitored first.
8. The energy storage power station fire warning system based on multi-parameter fusion according to claim 5 is characterized in that: The fire linkage mode of the linkage response module is as follows: based on the prediction results of the thermal runaway spread path, the fire extinguishing equipment within a range of 3 meters around the fire point is activated in a targeted manner to avoid secondary damage to equipment caused by full-area fire extinguishing; At the same time, parameter changes during the fire extinguishing process are recorded and fed back to the intelligent early warning module for model optimization.
9. The energy storage power station fire warning system based on multi-parameter fusion according to claim 5, characterized in that: The energy storage power station fire warning system also includes: a multi-cabin collaborative warning module; The multi-cabin collaborative early warning module automatically sends an early warning signal to the adjacent cabin when the corresponding prefabricated cabin triggers an alarm. The monitoring modules of the adjacent cabins increase the sampling frequency and activate the fire isolation doors between the cabins to block the fire spread channel.
10. The energy storage power station fire warning system based on multi-parameter fusion according to claim 5, characterized in that: The data fusion processing module supports data interaction with the energy storage power station's power management system, integrating operating data related to charge and discharge power and cycle number. When the charge and discharge rate is greater than 1C, it automatically increases the monitoring weight of voltage fluctuations and temperature gradients to adapt to the early warning needs under high-load conditions.
Citation Information
Cited By
Power battery thermal runaway early warning method and system based on multiple physical fields
CN120870938A
Method and system for early warning of thermal runaway of power battery based on multiple physical fields
CN120870938B
Thermal runaway management method and system for energy storage battery
CN121192329A
A method and system for thermal runaway management of energy storage batteries
CN121192329B
Intelligent safety protection system and method for power supply charging of underground coal mine power lithium battery and computer readable storage medium
CN121546773A
