A method and system for monitoring hydrogen sulfide leak anomalies

Abstract

The application discloses a hydrogen sulfide leakage anomaly monitoring method and system and relates to the technical field of gas monitoring. The hydrogen sulfide leakage anomaly monitoring system comprises a data monitoring module, an airflow correction module, a twin inversion module, a correlation judgment module, a leakage confirmation module, a linkage control module and a monitoring updating module. The application corrects the hydrogen sulfide concentration by constructing an airflow vector model and calculating a gas diffusion disturbance coefficient, so that the monitoring result is closer to the real distribution of the gas, and the spatial accuracy and reliability of the leakage detection are improved. The application introduces the corrected concentration data into a coupling domain for numerical evolution by establishing a double-physical-field digital twin model of a battery thermal field and an airflow field, so that the spatial inversion positioning and diffusion path prediction of the hydrogen sulfide concentration anomaly are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas monitoring, and in particular to a hydrogen sulfide leakage anomaly monitoring method and system. BACKGROUND

[0002] In the long-term operation process of lead-acid energy storage power stations, since the electrolyte is a dilute sulfuric acid solution, side reactions occur during the charging and discharging and aging stages of the battery, generating a small amount of hydrogen sulfide gas. When the battery pack is overcharged, the electrolyte is contaminated, or the pole plate is sulfided, the release amount of hydrogen sulfide will significantly increase. Hydrogen sulfide has high toxicity, strong corrosiveness and flammability, and is extremely easy to accumulate in a closed or semi-closed energy storage cabin environment, forming a safety hazard. In order to ensure the safe operation of lead-acid energy storage power stations, it is necessary to monitor and identify the hydrogen sulfide concentration in real time. However, the internal structure of the lead-acid energy storage cabin is complex, and the gas diffusion is affected by many factors such as ventilation state, temperature distribution and battery layout in the cabin, resulting in deviations in the measurement results of single-point sensors, and it is difficult to accurately reflect the actual leakage distribution and risk level.

[0003] The existing energy storage system safety monitoring mainly relies on a single gas sensor and a fixed threshold alarm mechanism to judge the anomaly by detecting the concentration changes of hydrogen sulfide or oxygen. Although this method can issue an alarm at a high concentration stage, it is slow to respond to small leaks in the early stages of gas diffusion, and is easily affected by air flow disturbances, changes in ventilation equipment operation and fluctuations in temperature and humidity, resulting in false positives and false negatives. In addition, the existing system usually lacks the ability to spatially invert the gas diffusion path and source location, and cannot accurately assess the risk level and linkage control strategy, nor can it consider the dynamic influence of battery health status and environmental changes on leakage risk. Therefore, traditional monitoring methods cannot meet the needs of lead-acid energy storage power stations for high-precision and intelligent monitoring of hydrogen sulfide leakage under complex operating conditions. SUMMARY

[0004] The present application proposes a hydrogen sulfide leakage anomaly monitoring method for lead-acid energy storage power stations to solve the problem of high-precision identification and spatial positioning of hydrogen sulfide leakage in air flow disturbance and multi-source coupled environments.

[0005] A hydrogen sulfide leakage anomaly monitoring method, comprising:

[0006] Real-time collection of data from each energy storage cabin monitoring point of the energy storage power station, including environmental parameters, ventilation equipment status, concentrations of hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride, and synchronous collection of battery health status parameters;

[0007] Obtaining air flow parameters according to environmental parameters and ventilation equipment status, obtaining a geometric structure model of the energy storage cabin, selecting an air flow vector model, calculating a gas diffusion disturbance coefficient, correcting the hydrogen sulfide concentration in the air flow field, and obtaining a corrected concentration;

[0008] The battery thermal distribution is obtained based on the environmental parameters and the health state parameters, a double-physical-field digital twin model is established by combining the corrected concentration, airflow parameters and a geometric structure model, and the spatial position of the concentration anomaly, the diffusion path and the concentration distribution of hydrogen sulfide at a preset time node are determined through inversion calculation;

[0009] Based on the corrected concentration, and the concentrations of hydrogen, carbon monoxide and hydrogen fluoride, the corresponding concentration change rate curves are constructed, the time sequence correlation and proportional relationship between the hydrogen sulfide concentration change rate curve and the rest concentration change rate curves are analyzed, and the fault type of the concentration anomaly is determined based on a battery fault mode library;

[0010] Based on the health state parameters, the environmental parameters and the fault type, the dynamic threshold model is adjusted, the corrected concentration and the hydrogen sulfide concentration change rate curve are monitored through the dynamic threshold model, and the leakage anomaly is determined;

[0011] According to the health state parameters, the corresponding grades of the corrected concentration, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution and the fault type, a comprehensive risk index is calculated, and a hierarchical linkage control strategy is executed based on the comprehensive risk index;

[0012] Based on the historical distribution of the concentration anomaly and the leakage anomaly, the monitoring coverage is evaluated, and the layout of the monitoring points is updated.

[0013] As a preferred technical solution of the present application, the calculation of the gas diffusion disturbance coefficient includes: the airflow parameters are composed of the monitoring point wind speed vector, the air volume distribution and the ventilation equipment state; according to the geometric structure model and the typical operation condition of the energy storage cabin, a basic airflow vector model inside the energy storage cabin is constructed offline by using the computational fluid dynamics method, and a flow field feature database under different ventilation equipment states is generated; the current airflow parameters are compared with the flow field feature database, the corresponding airflow vector model is selected in the basic airflow vector model, the turbulent kinetic energy distribution and the local vector of each region inside the energy storage cabin are obtained through the airflow vector model, the local vector is differentially calculated with the measured wind speed vector obtained by the monitoring point to form an airflow deviation field; based on the airflow deviation field and the distribution of the turbulent kinetic energy, the least square fitting algorithm is used to calculate the gas diffusion disturbance coefficient.

[0014] As a preferred technical solution of the present application, the airflow field correction of the hydrogen sulfide concentration includes: the measured hydrogen sulfide concentration of the monitoring point is spatially corresponding to the airflow vector model to obtain the gas diffusion disturbance coefficient of the monitoring point; based on the gas diffusion disturbance coefficient and the wind speed vector, the concentration drift deviation of the hydrogen sulfide concentration measurement caused by the airflow is calculated; the measured hydrogen sulfide concentration is additively compensated and corrected through the concentration drift deviation to obtain the corrected concentration.

[0015] As a preferred technical solution of the present application, the digital twin model of the double physical field comprises: obtaining the battery heat distribution through the temperature and health state parameters in the environmental data, the health state parameters including the charge-discharge rate, voltage, voltage consistency, internal resistance and historical charge-discharge cycle number; coupling the battery heat distribution, airflow parameter and geometric structure model to construct a double physical field coupling domain inside the energy storage cabin; in the double physical field coupling domain, a coupling model of the heat field equation, flow field equation and concentration equation is established by using the finite volume method, and the battery module heat dissipation channel and ventilation opening are introduced as boundary conditions; the modified concentration is introduced into the coupling model as a passive scalar to realize the synchronous evolution simulation of the heat field, flow field and concentration field, and form the digital twin model of the double physical field.

[0016] As a preferred technical solution of the present application, the inversion calculation comprises: inputting the environmental parameters and health state parameters in the digital twin model, performing numerical solution through the coupling model to obtain the distribution of airflow vector inside the energy storage cabin and the predicted concentration of hydrogen sulfide; performing difference calculation on the predicted concentration of hydrogen sulfide and the modified concentration, determining the spatial position of the concentration anomaly based on the spatial residual distribution of the two; calculating the gas transport velocity field according to the distribution of airflow vector and the turbulent kinetic energy distribution; solving the non-steady-state concentration transport equation in the transport velocity field to obtain the concentration distribution of hydrogen sulfide with time evolution; generating the diffusion path of hydrogen sulfide in the preset time window by spatial fitting of the concentration isosurface at different times.

[0017] As a preferred technical solution of the present application, the fault type of the concentration anomaly comprises: extracting feature parameters including peak occurrence time, rising rate and change duration based on each concentration change rate curve; obtaining the joint feature vector of multiple gases by analyzing the time sequence correlation and concentration proportion relationship between the hydrogen sulfide concentration change rate curve and the hydrogen, carbon monoxide and hydrogen fluoride concentration change rate curves; matching the joint feature vector with the pre-established battery fault mode library to determine the fault type.

[0018] As a preferred technical solution of the present application, the dynamic threshold model comprises: obtaining the battery health state deviation, environmental disturbance coefficient and fault risk weight based on the health state parameters, environmental parameters and fault type of the battery respectively, performing normalized weighted calculation on the battery health state deviation, environmental disturbance coefficient and fault risk weight to obtain a threshold adjustment factor; correcting the concentration threshold and change rate threshold according to the threshold adjustment factor, and the dynamic threshold model realizes adaptive update through time weighting and historical data backtracking algorithm.

[0019] As a preferred technical solution of the present application, the hierarchical linkage control strategy includes: according to the risk level corresponding to the comprehensive risk index, the hydrogen sulfide leakage event is divided into early warning level, alert level and emergency level; when the risk level is early warning level, local ventilation control is executed and risk prompt information is uploaded; when the risk level is alert level, the independent exhaust device of the leakage area is started, and the operation and maintenance system is notified to enter the monitoring intensive mode; when the risk level is emergency level, the energy storage cabin emergency linkage control action is executed, including power cut-off and ventilation fire extinguishing start; and the comprehensive risk index is formed by weighting the concentration weight, the electrical risk weight and the diffusion risk weight.

[0020] As a preferred technical solution of the present application, the updating of the layout of the monitoring points includes: collecting and storing historical distribution data of concentration anomalies and leakage anomalies, performing statistics according to spatial coordinates and occurrence frequencies, mapping the historical distribution data to the monitoring area defined by the geometric structure model of the energy storage cabin, and calculating a monitoring coverage index according to the spatial density of the corresponding anomalies of the historical distribution data and the spatial distance from the monitoring points to the anomaly positions; identifying areas with insufficient monitoring coverage, determining the position adjustment amount of the monitoring points based on spatial reconstruction and reachability analysis, generating coordinate suggestions for the migration or addition of the monitoring points, and updating the layout configuration of the monitoring points.

[0021] A hydrogen sulfide leakage anomaly monitoring system, comprising:

[0022] The data monitoring module: real-time collection of data of each energy storage cabin monitoring point of the energy storage power station, including environmental parameters, ventilation equipment state, concentrations of hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride, and synchronous collection of battery health state parameters;

[0023] The airflow correction module: obtains airflow parameters according to environmental parameters and ventilation equipment state, obtains the geometric structure model of the energy storage cabin, selects an airflow vector model, calculates a gas diffusion disturbance coefficient, corrects the hydrogen sulfide concentration in the airflow field, and obtains a corrected concentration;

[0024] The twin inversion module: obtains the battery heat distribution based on the environmental parameters and the health state parameters, establishes a digital twin model of double physical fields by combining the corrected concentration, the airflow parameters and the geometric structure model, determines the spatial position of the concentration anomaly, and the diffusion path and concentration distribution of hydrogen sulfide at a preset time node through inversion calculation;

[0025] The correlation judgment module: constructs corresponding concentration change rate curves based on the corrected concentration and the concentrations of hydrogen, carbon monoxide and hydrogen fluoride, analyzes the time sequence correlation and proportional relationship between the hydrogen sulfide concentration change rate curve and the remaining concentration change rate curves, and judges the fault type of the concentration anomaly based on a battery fault mode library;

[0026] Leakage confirmation module: based on health state parameters, environmental parameters and fault types, adjusting the dynamic threshold model, monitoring the corrected concentration and hydrogen sulfide concentration rate of change curve through the dynamic threshold model, determining the leakage anomaly;

[0027] Linkage control module: according to the health state parameters, the grade corresponding to the corrected concentration, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution and the fault type, the comprehensive risk index is calculated, and the hierarchical linkage control strategy is executed based on the comprehensive risk index;

[0028] Monitoring update module: based on the historical distribution of concentration anomalies and leakage anomalies, the monitoring coverage is evaluated, and the position of the monitoring point is updated.

[0029] The present application has the following advantages:

[0030] The present application overcomes the concentration drift problem caused by the complex air flow distribution and ventilation working condition change in the lead-acid energy storage cabin by constructing an air flow vector model and calculating the gas diffusion disturbance coefficient, and correcting the hydrogen sulfide concentration, so that the monitoring result is closer to the real distribution of the gas, and the spatial accuracy and reliability of the leakage detection are improved.

[0031] The present application realizes the spatial inversion positioning and diffusion path prediction of the hydrogen sulfide concentration anomaly by establishing a double-physical-field digital twin model of the battery thermal field and the air flow field, introducing the corrected concentration data into the coupling domain for numerical evolution, and directly determining the leakage source position and diffusion trend.

[0032] The present application realizes the fault type discrimination under the joint characteristics of multiple gases by constructing multiple gas concentration rate curves and analyzing the time sequence correlation and proportional relationship of hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride, and matching with the battery fault mode library, effectively distinguishing different leakage causes, and improving the pertinence and accuracy of diagnosis.

[0033] The present application overcomes the problem that the traditional fixed threshold method is not sensitive to the change of operating conditions by introducing a dynamic threshold model and adaptively adjusting the alarm threshold according to the battery health state, environmental disturbance and fault risk weight, realizes the dynamic optimization of the threshold with the operating state, and improves the sensitivity and robustness of the monitoring system.

[0034] The present application realizes the adaptive evolution and dynamic adjustment of the monitoring system by evaluating the monitoring coverage based on the historical concentration anomaly and leakage event distribution data, and optimizing the monitoring point layout combined with spatial reconstruction and reachability analysis, continuously improving the effective coverage rate and risk capture ability of the monitoring area. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to make the objects, technical solutions and advantages of the present application clearer, the following will further describe the present application with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the protection scope of the present application.

[0036] Figure 1 A structure schematic diagram of a hydrogen sulfide leakage anomaly monitoring system used in an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to make the objects, technical solutions and advantages of the present application clearer, the following will further describe the present application with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the protection scope of the present application.

[0038] Embodiment 1, a hydrogen sulfide leakage anomaly monitoring method, comprising the following steps:

[0039] Step S1: Real-time collection of data of each energy storage cabin monitoring point of the energy storage power station, including environmental parameters, ventilation equipment state, hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride concentration, and synchronous collection of battery health state parameters;

[0040] In an embodiment of the present application, a plurality of energy storage cabins are arranged in the energy storage power station, and a plurality of monitoring points are arranged in each energy storage cabin for collecting environmental data, gas concentration data and battery operation state information. The data of all the monitoring points are uniformly and synchronously transmitted to a central data processing unit through a data acquisition network to ensure that the time bases of various data are consistent.

[0041] The environmental parameters are used to reflect the thermal and humid states and ventilation characteristics of the inside and local area of the energy storage cabin, including temperature, humidity, air pressure, wind speed vector and air volume distribution data. The environmental temperature is collected by a digital temperature sensor, the humidity is obtained by a capacitive humidity sensor, and the air pressure is measured by a differential pressure sensor, and the data sampling interval is not more than 5 seconds. The wind speed vector is an important basic data for constructing the airflow parameter. The monitoring points are arranged at representative positions of the air inlet, air outlet and cabin passage, and the wind speed sensor of each monitoring point measures the three-dimensional components (velocity values along the cabin coordinate x, y and z directions) of the airflow in real time and generates the corresponding wind speed vector. The air volume distribution data is obtained by integrating the sectional flow rate of each ventilation air inlet, fan and ventilation pipeline, and is used to describe the gas transport capacity and spatial flow difference inside the cabin. The environmental parameters are used for subsequent airflow parameter calculation and airflow vector model matching, ventilation state identification, threshold model environmental correction and thermal field boundary condition setting.

[0042] The ventilation equipment status data includes operational information for ventilation fans, dampers, and cooling devices. This data is used to determine the operating condition of the ventilation equipment (e.g., normal ventilation, partial ventilation, or emergency exhaust) and provides input for subsequent selection of airflow vector models and calculation of gas diffusion disturbance coefficients. Monitoring content includes: fan start / stop status, operating mode (automatic, manual, or emergency), speed, damper opening, airflow output, filter pressure differential, and operating current. All ventilation equipment is equipped with an operational status detection module that uploads real-time operational status information at 2-second intervals.

[0043] Gas concentration data includes hydrogen sulfide ( ),hydrogen( The instantaneous concentrations of hydrogen sulfide (H2SO4), carbon monoxide (CO), and hydrogen fluoride (HF) were monitored. Hydrogen sulfide concentration was the primary monitoring target in this invention. Hydrogen sulfide is generated during thermal instability or side reactions in sulfur-containing energy storage batteries (such as sodium-sulfur batteries and sulfide solid-state batteries). It is highly toxic and corrosive, and its early detection is crucial for preventing leaks and ensuring personnel safety. Monitoring data for hydrogen, carbon monoxide, and hydrogen fluoride were used to establish a correlation model of multi-gas concentration change rates to determine the presence of fault symptoms related to electrochemical reactions.

[0044] The energy storage chamber is equipped with a multi-point composite gas sensor array, with each sensor in the array having independent spatial coordinates and calibration information. The sensors utilize electrochemical detection elements with high sensitivity and low cross-interference characteristics, and their range and resolution are as follows: ( ) (0–100ppm, 0.1ppm), ( (0–1000ppm, 1ppm), (C0) (0–100ppm, 0.1ppm), (HF) (0–10ppm, 0.05ppm). The sampling period is 1 second, and the collected data is uploaded after local filtering and outlier detection.

[0045] Battery health parameters are provided in real-time by the battery management system (BMS) of the energy storage system. These parameters include charge / discharge rate, voltage, current, internal resistance, consistency indicators (maximum voltage difference between cells), temperature distribution, state parameters (SoC, SoH), and historical cycle count. These parameters are used to estimate battery heat source distribution, determine operating conditions, and construct thermal boundary conditions for the dual-physics coupling domain. The sampling interval for health parameters is no more than 1 second, and they share a unified timestamp with gas and environmental data. Battery health parameters undergo multiple filtering and anomaly verification algorithms by the BMS before acquisition. Temperature distribution data is obtained through an array of thermistors within the module and is used to compare with ambient temperature to assess the degree of thermal imbalance.

[0046] This application is in monitoring At the same time, build , the concentration rate of change curve of CO, HF, and calculate its time sequence correlation and proportional relationship with the concentration rate of change curve, form a joint feature vector for fault type discrimination. In this way, different causes (such as overcharge side reaction, electrolyte / material decomposition, sealing failure, etc.) can be distinguished, and false positives and false negatives caused by a single threshold can be avoided.

[0047] and the correlation of Physical / chemical correlation (common sense in the art): overcharge or local polarization will enhance the hydrogen evolution reaction ( In sulfur-containing systems or lead-acid environments, the side reactions of the electrode / electrolyte, contaminants, or local sulfuration process may occur simultaneously or secondarily If rises before or simultaneously with , and the rate of change is positively correlated, it may indicate an electrochemically induced leakage / side reaction path rather than a pure diffusion disturbance. The "rate of change + time sequence correlation + proportional relationship" is used to confirm this coupling rather than accidental coincidence.

[0048] The correlation of C0 and Physical / chemical correlation (common sense in the art): C0 is common in the pyrolysis or local overheating oxidation of organic components / separators / adhesives. If C0 and appear in the same direction or near time lag, it suggests a thermal-chemical composite anomaly (such as a hot spot causing material decomposition, triggering a sulfur-containing component / contaminant reaction to produce ). The time difference between the two gas rate peaks and the rise rate ratio are used to quantify the "thermal-chemical" linkage feature.

[0049] The correlation of HF and Physical / chemical correlation (common sense in the art): HF is the decomposition or acidification product of fluorine-containing materials / components (such as fluoropolymer seals / insulation, coatings; and fluorine-containing electrolytes when mixed or arranged in the same station with lithium-based equipment) under high temperature / strong acid / hydrolysis conditions. If HF and exist synchronously or in short time lag: indicating an acid-thermal coupling scenario: local acidity / temperature rise causes fluorine-containing materials to decompose (HF ), while sulfur-containing components / contaminants undergo side reactions ( For lead-acid systems only, HF often indicates that fluorine-containing components of the cabin or accessories are degraded by heat or acid; if there are lithium-based equipment in the same station or adjacent area, HF comes from the hydrolysis side product of its electrolyte, which is a "cross-equipment gas disturbance" signal and should be investigated jointly. This application explicitly includes HF in the multi-gas rate and proportional analysis framework as one of the feature gases to participate in fault type discrimination.

[0050] Note: The upper chemical mechanism is a general engineering principle; when the algorithm of the present application is applied, the time sequence correlation and proportional relationship of the change rate (rather than a single chemical hypothesis) are used as the basis for judgment to ensure objectivity and verifiability.

[0051] Health parameters such as charge / discharge rate, voltage, current, internal resistance, consistency (maximum pressure difference of single cell), temperature distribution, SoC, SoH, cycle number, etc. are obtained from BMS, which are used to: calculate heat source distribution and construct double physical field coupling domain (heat-flow-concentration) with environment / airflow; provide real working condition boundary for digital twin inversion, so as to more accurately locate the abnormal concentration spatial position and diffusion path; as an important input of dynamic threshold model, adaptively adjust the threshold with health / environment / fault risk weight, reduce false alarm and missed alarm. The above uses are respectively described in the health parameter set and use description, double physical field twin / inversion, dynamic threshold model clauses.

[0052] Correlation logic (method of the present application + engineering mechanism): overcharge / poor consistency / internal resistance increase Local heating and polarization enhancement Hydrogen evolution and sulfur-containing side reaction probability More likely to occur Early weak leakage. Uneven temperature distribution / hot spot Thermal decomposition of materials / seals (may see CO, HF features) and drive Synchronous or secondary rise. These states are mapped to heat-flow conditions through the twin model, changing the transport and diffusion form, and then affecting The spatial distribution and time sequence of detection.

[0053] Multi-gas— The criteria and patterns associated are shown in Table 1:

[0054] Table 1:

[0055]

[0056] The statistical threshold in the above table is the value taken by the present embodiment, and the algorithm level is determined by "change rate curve, time sequence correlation, proportional relationship + pattern library matching", which is not limited to a fixed threshold.

[0057] Step S2: Obtain airflow parameters according to environmental parameters and ventilation equipment state, obtain geometric structure model of energy storage cabin, select airflow vector model, calculate gas diffusion disturbance coefficient, correct hydrogen sulfide concentration based on airflow field, and obtain corrected concentration;

[0058] The internal airflow of the energy storage cabin is complex, and the ventilation working condition changes, resulting in "drift" of the measured concentration. In order to restore the measured concentration to be closer to the real spatial distribution, a gas diffusion disturbance coefficient is introduced, and based on the airflow vector model, The concentration is corrected by applying airflow field correction to obtain the corrected concentration. The corrected concentration serves as a unified physical background quantity for subsequent twin inversion and risk assessment, significantly improving the reliability of localization and diffusion prediction.

[0059] In the actual testing process, A: directly use the actual measurements. The comparison of the monitoring results of concentration (without introducing the gas diffusion disturbance coefficient) and B: after additive compensation correction according to the airflow vector model (introducing the gas diffusion disturbance coefficient) is shown in Table 2.

[0060] Table 2:

[0061]

[0062] The calculation of the gas diffusion disturbance coefficient includes: airflow parameters consisting of wind speed vectors at monitoring points, air volume distribution, and ventilation equipment status; based on the geometric structure model of the energy storage chamber and typical operating conditions, a basic airflow vector model inside the energy storage chamber is constructed offline using computational fluid dynamics methods, and a flow field characteristic database under different ventilation equipment statuses is generated; the current airflow parameters are compared with the flow field characteristic database, and the corresponding airflow vector model is selected from the basic airflow vector model; the turbulent kinetic energy distribution and local vectors of each region inside the energy storage chamber are obtained through the airflow vector model; the local vectors are compared with the measured wind speed vectors obtained from the monitoring points to form an airflow deviation field; based on the airflow deviation field and the distribution of turbulent kinetic energy, the gas diffusion disturbance coefficient is calculated using a least squares fitting algorithm; the gas diffusion disturbance coefficient is used to quantitatively characterize the degree of deviation between the actual flow field and the simulated flow field inside the energy storage chamber.

[0063] The process of correcting the hydrogen sulfide concentration by adjusting the airflow field includes: spatially mapping the measured hydrogen sulfide concentration at the monitoring point to the airflow vector model to obtain the gas diffusion disturbance coefficient at the monitoring point; calculating the concentration drift deviation caused by the airflow to the hydrogen sulfide concentration measurement based on the gas diffusion disturbance coefficient and the wind speed vector; and performing additive compensation correction on the measured hydrogen sulfide concentration using the concentration drift deviation to obtain the corrected concentration.

[0064] In one embodiment of the present invention, airflow parameters are used to characterize the directionality and velocity distribution of airflow inside the energy storage chamber, serving as key inputs for subsequent gas diffusion modeling and concentration correction. The airflow parameters are constructed from the environmental data collected in step S1 and the status of the ventilation equipment, including three parts: the wind speed vector at the monitoring point, the air volume distribution, and the operating status of the ventilation system. The environmental data provides the actually measured wind speed vector value, while the ventilation equipment status provides information such as fan speed, valve opening, and ventilation mode. The combination of these two types of data reflects the real-time characteristics of airflow within the chamber.

[0065] The geometric model of the energy storage cabin is derived from a three-dimensional digital model in the cabin design stage, which records geometric information such as the shape of the cabin, the ventilation duct, the arrangement of the battery module, the ventilation air outlet, and the installation position of the equipment. Based on the model, the calculation domain division of the cabin interior is established for the subsequent spatial mapping of the airflow vector model and the flow field characteristic database.

[0066] The airflow vector model is used to describe the direction and speed of gas flow at each position inside the energy storage cabin. This model is not a single fixed field, but a set of typical flow field samples established in advance according to different ventilation states, equipment operating conditions, and external environmental conditions. Before the deployment of the monitoring system, the computational fluid dynamics (CFD) method is used to generate the basic airflow vector model offline, and it is solved under multiple operating modes to form a flow field characteristic database. The flow field characteristic database is indexed by the ventilation equipment state (such as the number of fans, speed, and air valve opening) and the external environmental boundary conditions, supporting subsequent model matching and disturbance correction.

[0067] The gas diffusion disturbance coefficient is used to quantitatively characterize the deviation between the actual airflow state inside the energy storage cabin and the offline simulated flow field, which is an innovative parameter proposed by the present invention. This coefficient reflects the influence of flow field disturbances caused by equipment operation fluctuations, local shielding, and cabin door opening on gas diffusion characteristics. Its physical meaning is the correction proportion of gas diffusion capacity relative to the reference flow field, which is used to correct the measured results of hydrogen sulfide concentration at each monitoring point.

[0068] Specifically, the current collected airflow parameters are compared with the flow field characteristic database, and the basic airflow vector model that best matches the current ventilation state is selected. By calculating the difference between the local flow velocity vector in the model and the measured wind speed vector at the monitoring point, a gas flow deviation field is formed. The gas flow deviation field is used to describe the local deviation trend of the actual flow relative to the simulated reference field. Combined with the distribution of turbulent kinetic energy in the flow field, a least squares fitting algorithm is used to obtain the gas diffusion disturbance coefficient of each monitoring point and its neighborhood (the least squares algorithm minimizes the sum of squares of local deviations to ensure that the correction effect of the coefficient on the entire monitoring domain is optimally fitted). The gas diffusion disturbance coefficient is a dimensionless parameter, with a value range of 0 to 1.5. The closer the value is to 1, the higher the consistency between the actual flow field and the simulated flow field, and the value deviating from 1 indicates that there is a significant ventilation disturbance. The sliding time window and weight smoothing mechanism are introduced in the calculation process, making the disturbance coefficient change continuously in the time dimension, ensuring the stability and traceability of the correction process.

[0069] The air flow field correction is used to eliminate the deviation of hydrogen sulfide concentration measurement caused by the change of ventilation state. Specifically, the measured hydrogen sulfide concentration value of the monitoring point is corresponded with the selected air flow vector model in space position, and the gas diffusion disturbance coefficient of the corresponding position is extracted; according to the wind speed vector of the monitoring point and the disturbance coefficient, the influence deviation of the air flow on the concentration measurement is calculated, which represents the concentration drift caused by the directionality of wind speed and the intensity of turbulence; the concentration drift deviation is used as the correction factor to additively compensate the measured hydrogen sulfide concentration, and the concentration value after the air flow field correction is obtained, which is recorded as the corrected concentration. The correction process is logically equivalent to field domain normalization of the concentration signal measured by each monitoring point, so that the measurement results under different positions and different ventilation conditions can be compared in a unified physical background. The corrected concentration reflects the real distribution level of hydrogen sulfide under the current air flow condition, and provides the initial concentration boundary condition for the subsequent establishment of the digital twin model coupled with heat-flow-concentration.

[0070] The introduction of the gas diffusion disturbance coefficient avoids the misjudgment problem caused by ignoring the local air flow difference in the traditional method, so that the monitoring result more accurately reflects the leakage source strength and spatial diffusion trend. The value of the coefficient is not artificially set, but automatically calculated by comparing the difference between the measured wind speed and the simulation model. The establishment and update of the basic air flow vector model are based on the actual energy storage cabin geometry and ventilation working condition data, which are generated by combining numerical simulation and experimental verification, ensuring the physical authenticity. In engineering application, the air flow field correction process is executed on the edge computing node with a period of seconds, ensuring the real-time and continuity of the corrected concentration. After time synchronization, the correction result is transmitted to the double-physical-field digital twin model in step S3, which is used to build the initial field of concentration distribution, so as to realize accurate simulation of the diffusion behavior of hydrogen sulfide.

[0071] Step S3: obtaining the battery heat distribution based on the environmental parameters and health state parameters, combining the corrected concentration, air flow parameters and geometric structure model, establishing a double-physical-field digital twin model, determining the spatial position of concentration anomaly and the diffusion path and concentration distribution of hydrogen sulfide at the preset time node through inversion calculation;

[0072] The establishment of the double-physical-field digital twin model includes: obtaining the battery heat distribution through the temperature in the environmental data and the health state parameters, the health state parameters including the charge-discharge rate, voltage, voltage consistency, internal resistance and historical charge-discharge cycle number; coupling the battery heat distribution, air flow parameters and geometric structure model to build a double-physical-field coupling domain inside the energy storage cabin; in the double-physical-field coupling domain, a coupling model of heat field equation, flow field equation and concentration equation is established by using the finite volume method, and the battery module heat dissipation channel and ventilation air inlet are introduced as boundary conditions; the corrected concentration is introduced into the coupling model as a passive scalar to realize the synchronous evolution simulation of the heat field, flow field and concentration field, and form a double-physical-field digital twin model.

[0073] The inversion calculation comprises: inputting environmental parameters and health state parameters in the digital twin model, performing numerical solution through the coupling model, obtaining the distribution of airflow vector inside the energy storage cabin and the predicted concentration of hydrogen sulfide; performing difference calculation on the predicted concentration of hydrogen sulfide and the corrected concentration, determining the spatial position of the concentration anomaly based on the spatial residual distribution of the two, and adopting the least mean square error inversion algorithm; calculating the gas transport velocity field according to the distribution of the airflow vector and the turbulent kinetic energy distribution; solving the non-steady-state concentration transport equation in the transport velocity field to obtain the concentration distribution of hydrogen sulfide over time; and generating the diffusion path of hydrogen sulfide within a preset time window by spatial fitting of the concentration isosurface at different times.

[0074] The establishment of the double-physical-field digital twin model relies on the real structure of the energy storage cabin, the measurement point layout and the ventilation working condition data, all inputs come from field measurement, which ensures that the model has physical authenticity and verifiability. The model adopts the finite volume method for discretization and time step solution, and the calculation is stable and has good convergence. The inversion algorithm takes data-driven as the core, minimizes the difference between the measured and predicted values through continuous optimization, avoiding the subjectivity of manually specifying the leakage source or threshold. The diffusion path result generated by the model is verified by historical data, providing spatial range and time warning basis for emergency linkage control.

[0075] In an embodiment of the present application, the battery thermal distribution is obtained based on the combination of environmental parameters and health state parameters. The battery thermal distribution is the basic input for the calculation of the internal thermal field of the energy storage cabin, which is used to describe the heat generation intensity of the battery module and the heat exchange state with the surrounding air. The thermal distribution is determined by the internal heat generation of the battery and the external environmental heat exchange condition: the equivalent heat generation power of each battery module is calculated according to the health state parameters (including charge-discharge rate, voltage, current, internal resistance, consistency index and historical cycle number), forming the internal heat source distribution; the air side heat exchange boundary and heat dissipation channel condition are determined according to the environmental parameters and the state of the ventilation equipment (including temperature, humidity, wind speed vector, air volume distribution, air opening degree and ventilation mode); the heat generation distribution and the heat exchange boundary are mapped into the spatial domain of the geometric structure model of the energy storage cabin together to form a complete battery thermal distribution field. The battery thermal distribution field reflects the temperature variation characteristics of each position inside the cabin body, and is the basic input for the subsequent solution of the thermal field equation. Among them, the wind speed vector and the air volume distribution are directly used to calculate the convective heat transfer coefficient, which reflects the influence of the ventilation state on the heat transfer efficiency. Through the above joint calculation, the heat source distribution data with physical consistency and environmental response characteristics are obtained.

[0076] The battery heat distribution in the present embodiment refers to a three-dimensional temperature field formed by the heat generated by the battery and air convection in the energy storage cabin. The establishment of the distribution relies on two types of data, namely, the measured health state parameters and environmental parameters. The former determines the heat source strength, and the latter determines the heat dissipation conditions. The two are coupled to form a complete temperature spatial distribution. This combined modeling method ensures the consistency of the heat field input and the ventilation state, avoids the deviation in heat estimation caused by ignoring the environmental convection effect, and improves the authenticity and reproducibility of the digital twin model.

[0077] On this basis, a double-physical-field digital twin model is established by combining the corrected concentration, airflow parameters and energy storage cabin geometric structure model. The model realizes the coupled simulation of temperature, velocity and gas concentration in the cabin by simultaneously solving the heat field and flow field in the unified calculation domain. Among them, the heat field is provided with source term input by the battery heat distribution, the flow field is provided with boundary driving conditions by the airflow parameters, and the corrected concentration participates in the field evolution as a passive scalar.

[0078] The geometric structure model records the spatial layout, air duct structure and ventilation opening position inside the energy storage cabin. The airflow parameters come from the real-time wind speed vector data and gas diffusion disturbance coefficient corrected in step S2. Through the coordinate mapping relationship, the airflow parameters are projected into the calculation domain defined by the geometric structure model to form a spatially continuous airflow distribution field.

[0079] The double-physical-field coupling domain refers to the region inside the energy storage cabin that is affected by both thermal convection and ventilation disturbance. In this domain, air temperature, flow velocity and gas concentration are interdependent: temperature gradient affects air density and buoyancy, thereby changing the local airflow distribution; airflow determines the direction and speed of gas diffusion. Therefore, the coupling relationship between the heat field and the flow field is established by the finite volume method, and on this basis, the dynamic evolution equation of the concentration field is introduced. All equations are discretized using the same time step and spatial grid, ensuring the consistency and stability of the calculation.

[0080] The corrected concentration obtained in step S2 is introduced into the model as a passive scalar, representing the real distribution state of the gas at the initial time. The passive scalar does not affect the momentum equation of the fluid during the solving process, but is updated synchronously with the flow field and diffusion coefficient. This setting ensures that the evolution of the concentration is synchronized with the changes in temperature and airflow, realizing the digital mapping of the real cabin conditions.

[0081] The "double-physical-field digital twin model" described in the present embodiment refers to a calculation model that dynamically simulates the heat-flow-concentration three-field coupling evolution process inside the energy storage cabin in a virtual space using actual environmental data, airflow parameters and battery heat source data. Its core features include: real-time synchronization with measured data; one-to-one mapping of actual operating conditions; automatic updating of heat source and flow field boundary conditions in each calculation period to dynamically track the gas diffusion state in the cabin.

[0082] The objective of the inversion calculation is to determine the location of the hydrogen sulfide anomaly source, the diffusion direction, and the concentration distribution change over time based on the difference between the simulation results of the digital twin model and the measured corrected concentration.

[0083] The environmental parameters, health status parameters, and corrected concentration at the current time are input into the digital twin model, and the in-cabin airflow vector distribution and predicted concentration distribution are calculated through the coupling solver module. The predicted concentration reflects the theoretical diffusion state of the gas under the current ventilation and thermal conditions. The predicted concentration and the measured corrected concentration are one-to-one corresponding in space, and the difference between the two is calculated to form a spatial residual field, which is used to represent the deviation between the measured data and the model expectation. Through the least mean square error optimization algorithm, the spatial coordinates of the residual concentration center are located, and they are determined as the possible location of the concentration anomaly. The algorithm considers the coverage range and the density of the measuring points in the spatial search process to ensure the physical reasonableness of the results.

[0084] Based on the airflow vector distribution and turbulent kinetic energy information in the digital twin model, the gas transport velocity field is calculated, which is used to describe the actual movement trend of hydrogen sulfide. Combined with this velocity field, the non-steady-state concentration transport equation is solved to obtain the concentration distribution data over time. At each preset time node, the concentration isosurface is extracted, and the three-dimensional diffusion path is generated through spatial fitting and interpolation methods. The diffusion path reflects the time sequence rule and the restricted direction of the gas diffusion from the anomaly source point to the surrounding space.

[0085] The inversion process completely depends on the difference between the measured data and the model output, rather than the artificially set source point hypothesis. The model automatically adjusts the anomaly source location in each iteration to minimize the difference between the predicted value and the measured value, thereby ensuring the statistical and physical reasonableness of the results. It avoids the problem of traditional single-point threshold alarm that cannot be located, and realizes the source inversion and dynamic tracking based on data consistency.

[0086] During operation, the digital twin model is updated at a fixed time step, and each cycle automatically loads new environmental parameters, health status parameters, and corrected concentration data. When a heat source change or a ventilation state mutation is detected, the recalculation of the local area grid is automatically triggered to keep the model synchronized with the actual state. The model output includes the following contents: the spatial coordinates and uncertainty range of the hydrogen sulfide anomaly; the diffusion path and its time sequence concentration distribution; the gas transport velocity field and local turbulent characteristics; the model residual and confidence score for subsequent risk assessment to quantify the reliability of the model. The above output data will be transmitted to the multi-gas concentration change rate analysis module in step S4 to establish the time sequence correlation between hydrogen sulfide and other gases, and to further identify the fault type.

[0087] Step S4: Based on the modified concentration, and the concentrations of hydrogen, carbon monoxide, hydrogen fluoride, a corresponding concentration rate of change curve is constructed, the time sequence correlation and proportional relationship of the hydrogen sulfide concentration rate of change curve and the rest of the concentration rate of change curve are analyzed, and the fault type of the concentration anomaly is judged based on a battery fault mode library;

[0088] The judgment of the fault type of the concentration anomaly comprises: based on each concentration rate of change curve, characteristic parameters including peak appearance time, rising rate and change duration are extracted; by analyzing the time sequence correlation and concentration proportional relationship between the hydrogen sulfide concentration rate of change curve and the hydrogen, carbon monoxide and hydrogen fluoride concentration rate of change curve, a joint feature vector of multiple gases is obtained; and the joint feature vector is matched with a pre-established battery fault mode library to determine the fault type.

[0089] The multi-gas coupling analysis breaks through the limitation of the traditional single-gas concentration threshold judgment, and identifies complex anomalies triggered by different chemical mechanisms; the dual features of rate of change and proportion make the algorithm have the ability to distinguish both the leakage intensity and the reaction type; and the mode library matching mechanism replaces the subjective threshold judgment, and realizes the fault type identification based on statistical rules.

[0090] In an embodiment of the present application, based on the modified concentration data, the change trends of hydrogen sulfide (H2S) ), hydrogen (H2) ), carbon monoxide (CO) and hydrogen fluoride (HF) in the same time sequence are dynamically calculated.

[0091] The concentration rate of change curve refers to the instantaneous rate of change curve of the concentration of each gas with time at a uniform time step. It is obtained by calculating the concentration increment ratio in a continuous sampling period, and a weighted smoothing algorithm is used to suppress transient noise, so that the result can truly reflect the concentration change trend. The time resolution of the rate of change is consistent with the sampling period (usually 1 second), and shares the time reference with the modified concentration.

[0092] The modified concentration comes from the gas flow field correction result of step S2; the original signals of the concentrations of each gas come from the multi-gas sensor array of step S1; the time alignment and noise elimination process is performed by a central data synchronization module, ensuring that the time sequences of different gases are completely consistent.

[0093] For each gas concentration rate of change curve, the following key features are automatically extracted: peak appearance time (indicating the time point at which the gas concentration rises to a peak); rising rate (reflecting the speed of gas generation or diffusion); duration (describing the length of time that the gas maintains in an abnormal state); and decay slope (used to judge the dissipation speed of the gas under ventilation conditions).

[0094] Temporal correlation is used to measure the temporal coupling relationship between changes in the concentrations of multiple gases and is a key technical aspect of this invention for joint determination of multiple gases. Temporal correlation is defined as the degree of synchronization and proportional matching of the rate of change curves of two or more gases on the time axis. If two curves exhibit similar rising or falling characteristics within similar time periods, a positive temporal correlation is considered to exist; if they exhibit opposite characteristics, a negative correlation is considered.

[0095] The weighted cross-correlation algorithm is used to calculate the temporal correlation coefficient between any two gases. The algorithm incorporates a sliding window and dynamic threshold constraints to prevent misjudgments caused by short-term fluctuations, ensuring the smoothness and stability of the results. When the correlation coefficient between hydrogen sulfide and hydrogen, carbon monoxide, or hydrogen fluoride exceeds a preset correlation level threshold (determined based on historical experimental statistics), it is considered that there is an abnormal coupling caused by electrochemical reactions or material decomposition during that period.

[0096] Calculate the relative proportion of the peak values ​​of the rate of change of each gas concentration, for example: and The rate of change can reflect the intensity of sulfide side reactions; The ratio with HF can indicate electrolyte decomposition or membrane material reaction; The ratio to CO can be used to distinguish between thermal runaway and overcharge leakage. By constructing a joint feature vector of multiple gases using the above proportional characteristics and correlation coefficients, this vector comprehensively describes the multi-gas response characteristics of the in-cabin chemical reaction in dimensions of time, rate of change, and proportional relationship.

[0097] In one embodiment of the present invention, single-gas monitoring and multi-gas monitoring were tested by experiments, and the comparison results are shown in Table 3.

[0098] Table 3:

[0099]

[0100] The battery fault mode library serves as the knowledge benchmark for identification, containing various typical fault characteristic data summarized from experiments and long-term operation. Each fault mode is stored in the form of a triplet: "gas change rate characteristic + health status parameter + environmental conditions," covering different types such as overcharge, over-discharge, internal short circuit, electrolyte decomposition, and seal leakage. Each item in the mode library is accompanied by time characteristics (peak delay, duration), gas ratio characteristics, and risk level labels.

[0101] The combined feature vector is input to the pattern library matching module, and the closest fault type is identified through a multi-dimensional similarity calculation method (composite measurement based on Euclidean distance and weighted cosine similarity). If the similarity of the matching result exceeds the confidence threshold (for example, ≥ 0.85), it is determined to be this type of fault; if it is lower than the threshold, it enters the extended search mode, compares adjacent class samples, and recalculates through a time sliding window.

[0102] The fault determination result includes: fault type (such as "electrolyte decomposition leakage", "overcharge side reaction", "local seal failure", etc.); matching confidence; corresponding multi-gas response mode description (including characteristic gas and time sequence); recommended risk level (used for step S5 dynamic threshold adjustment). The fault identification process is completely based on statistical matching of measured concentration change rate and pattern library data, without introducing any artificially set fixed threshold or empirical formula. All feature vectors come from the physical change law of measured data, so the determination has objective verifiability and reproducibility. The construction of the pattern library is based on real battery tests and historical operation records, ensuring the engineering applicability of the algorithm recognition.

[0103] The fault risk weight parameter in the fault type determination threshold correction is used to dynamically correct the alarm threshold of hydrogen sulfide concentration and change rate. The combined feature vector and matching confidence data are simultaneously written into the operation database for subsequent model retraining and monitoring point optimization analysis.

[0104] Step S5: Based on the health state parameters, environmental parameters and fault type, adjust the dynamic threshold model, monitor the corrected concentration and hydrogen sulfide concentration change rate curve through the dynamic threshold model, and determine the leakage anomaly;

[0105] To avoid false positives and late reports of fixed thresholds under different ventilation / load conditions, a dynamic threshold model is introduced, which integrates battery health state deviation, environmental disturbance and fault risk weight to adjust the concentration threshold and change rate threshold in real time. Its role is to balance sensitivity and stability under working condition switching and environmental fluctuation scenarios, and reduce false positives / misses.

[0106] In an embodiment of the present application, the fixed threshold and the dynamic threshold are tested respectively through experiments, and the comparison results are shown in Table 4.

[0107] Table 4:

[0108]

[0109] The adjustment dynamic threshold model comprises: obtaining a battery health state deviation, an environmental disturbance coefficient and a fault risk weight based on a health state parameter, an environmental parameter and a fault type of the battery respectively, performing normalized weighted calculation on the battery health state deviation, the environmental disturbance coefficient and the fault risk weight to obtain a threshold adjustment factor; and correcting a concentration threshold and a change rate threshold according to the threshold adjustment factor, wherein the dynamic threshold model is updated adaptively through a time weighting and a historical data backtracking algorithm.

[0110] The battery health state deviation is a difference between the health state parameter and a nominal health threshold, the environmental disturbance coefficient is calculated according to temperature, humidity and wind speed change rate, and the fault risk weight is determined according to a severity level corresponding to the identified fault type. All threshold corrections are generated based on measured parameters and statistical learning model calculation results, without using fixed empirical values or manual adjustment.

[0111] In an embodiment of the present application, the dynamic threshold model is used to adaptively adjust the threshold parameters of the leakage judgment under different operating environments and battery health states, thereby avoiding the misjudgment problem caused by the fixed threshold. The model calculates a threshold adjustment factor through a multi-factor weighting algorithm based on the real-time input battery health state parameters, environmental parameters and fault types, which is used to dynamically correct the concentration threshold and the change rate threshold of hydrogen sulfide. The core idea of the model is to change the concentration threshold in real time with the working condition based on the system operating state and the risk weight as the adjustment core. The dynamic threshold model does not rely on artificial experience setting, but realizes self-learning mechanism driven by measured data.

[0112] The health state parameters include voltage consistency, internal resistance, charge-discharge rate, temperature distribution and SoH (state of health). The battery health state deviation is calculated based on the difference between the above parameters and the nominal health threshold, which reflects the deviation degree of the current state of the battery relative to the ideal state. For example, when the voltage consistency deviation exceeds the standard value or the temperature difference in the module increases, the deviation increases, and the threshold adjustment direction tends to tighten.

[0113] The environmental parameters include temperature, humidity, wind speed vector and air volume distribution. The environmental disturbance coefficient is calculated by analyzing the temperature and humidity change rate and the wind speed fluctuation amplitude. The environmental disturbance coefficient reflects the influence of the ventilation state and the external airflow change on the gas diffusion capacity. In a high-ventilation or high-disturbance environment, the gas dilution speed is fast, and the alarm threshold can be appropriately adjusted upward; in a low wind speed and closed condition, the diffusion capacity decreases, and the alarm threshold needs to be reduced.

[0114] A fault risk weight based on historical sample statistics is preset for each fault type. The risk weight is used to quantify the influence of different fault types on the severity of the leakage: electrolyte decomposition or membrane material reaction high risk weight; mild side reaction or over-discharge anomaly Medium risk weight; slight sealing leakage Low risk weight. All risk weights are derived from experimental data and statistical regression results of long-term running samples, not artificially set.

[0115] The above three types of input parameters are normalized to obtain X1 (health deviation amount), X2 (environmental disturbance coefficient), and X3 (fault risk weight). The threshold adjustment factor Ft is calculated by weighted combination: Ft = w1X1 + w2X2 + w3X3, wherein the weight coefficients w1, w2, and w3 are dynamically updated by the model self-learning module. The expression is self-adaptively corrected for weight values according to the early monitoring effect by a historical backtracking optimization algorithm.

[0116] The threshold adjustment factor Ft is used to dynamically update the initial reference of the reference threshold C0 (hydrogen sulfide concentration) and R0 (hydrogen sulfide change rate):

[0117] The corrected concentration threshold: Ct = C0 (1-kcFt); the corrected change rate threshold: Rt = R0 (1-krFt);

[0118] Wherein kc and kr are self-calibration coefficients in the model, used to prevent over-adjustment. When the environment or health state fluctuates, the threshold will be self-adaptively changed to maintain the balance between sensitivity and stability.

[0119] The dynamic threshold model includes a time weighting module and a historical backtracking algorithm: the time weighting module adjusts the threshold update frequency according to the data change speed, increases the response speed in the stage of rapid change, and reduces the fluctuation in the stable stage. The historical backtracking algorithm records the abnormal marks and threshold deviation trends in the past period of time, which is used to prevent continuous misjudgment or repeated alarms. The real-time self-adaptation and long-term stability of the system are ensured.

[0120] When the measured corrected concentration or concentration change rate continuously exceeds the corresponding dynamic threshold, and the duration exceeds the set minimum stable period (which is determined according to the monitoring history statistics), it is determined that there is a hydrogen sulfide leakage anomaly. In order to prevent false alarms triggered by single-point anomalies, a multi-point consistency confirmation mechanism is adopted: if adjacent monitoring points all appear over-threshold phenomenon in the same time window, it is confirmed that the leakage event is valid; if only a single point exceeds the threshold and the duration is shorter than the statistical stable period, it is marked as a "suspicious event" and does not trigger an alarm.

[0121] Step S6: According to the health state parameter, the level corresponding to the corrected concentration, the spatial position of the concentration anomaly, the diffusion path and concentration distribution, and the fault type, a comprehensive risk index is calculated, and a hierarchical linkage control strategy is executed based on the comprehensive risk index;

[0122] The parameters of the comprehensive risk index are obtained by actual measurement and algorithm self-learning by the system, and there is no artificial threshold setting.

[0123] The execution hierarchical linkage control strategy includes: according to the risk level corresponding to the comprehensive risk index, the hydrogen sulfide leakage event is divided into a warning level, an alert level and an emergency level; when the risk level is the warning level, local ventilation control is executed and risk prompt information is uploaded; when the risk level is the alert level, the independent exhaust device of the leakage area is started, and the operation and maintenance system is notified to enter the monitoring intensive mode; when the risk level is the emergency level, the energy storage cabin emergency linkage control action is executed, including power cut-off and ventilation fire extinguishing start; and the comprehensive risk index is formed by weighting the concentration weight, the electrical risk weight and the diffusion risk weight.

[0124] In an embodiment of the present application, the comprehensive risk index is used to quantify the overall danger degree of the hydrogen sulfide leakage event in the energy storage cabin, and reflects the comprehensive influence of gas concentration intensity, electrical risk and diffusion range. The index takes multi-source data fusion as the core, and realizes the unified quantification of different dimension risks through normalization and weighted calculation.

[0125] The calculation logic of the comprehensive risk index follows the following principles: the concentration level reflects the leakage intensity; the electrical level reflects the potential trigger source risk; and the diffusion level reflects the threat range of gas propagation in space. By comprehensively weighting the three types of risk factors, a risk index with engineering quantifiability, physical interpretability and time sequence traceability can be obtained.

[0126] The input data of the comprehensive risk index calculation is derived from the previous steps, including: the modified concentration and the level parameter, the modified concentration reflects the true concentration level of hydrogen sulfide under the current air flow condition; according to the safety standard (such as the occupational safety hydrogen sulfide limit) and the equipment corrosion resistance grade, the modified concentration is divided into multiple levels (such as low, medium, high and extremely high), and each level corresponds to different risk reference weight.

[0127] The spatial position of the concentration anomaly and the diffusion path, the spatial position determines the geometric coordinates of the abnormal source and the local risk aggregation area; the diffusion path and the concentration distribution describe the time sequence range and the potential influence domain of gas propagation. The wider the diffusion range and the greater the concentration gradient, the higher the diffusion risk weight.

[0128] The battery health state parameter and the fault type, the battery health state reflects the potential electrical safety hidden danger; the fault type is used to determine the possible gas generation mechanism and the persistence characteristics. For example: the electrolyte decomposition fault corresponds to a higher continuous release risk; the local sealing leakage corresponds to a limited risk.

[0129] The environmental parameters, including temperature, humidity, wind speed vector and air volume distribution, are used to modify the gas diffusion and dilution effect, and ensure the environmental adaptability of the risk index.

[0130] The comprehensive risk index Ra is composed of three types of weight factors: concentration weight Wc, representing the intensity risk of hydrogen sulfide; electrical risk weight We, representing the battery state and failure trigger potential; diffusion risk weight Wd, representing the influence range of the leakage space. All three are dimensionless quantities after standardization, with a value range of 0-1. Each weight value is determined by automatic learning of the model, not specified by humans.

[0131] The comprehensive risk index is calculated by multi-dimensional weighted fusion: Comprehensive risk index = (concentration level factor Concentration weight) + (electrical risk factor Electrical risk weight) + (diffusion range factor Diffusion risk weight); The concentration level factor is derived from the normalized value of the modified concentration level; The electrical risk factor is determined by the battery health deviation and the failure risk weight; The diffusion range factor is calculated according to the diffusion path length, abnormal volume and average concentration gradient. The model updates the three factors and weights in real time at each sampling period, and corrects the coefficients according to historical statistical feedback to ensure that the output results are smooth and continuous.

[0132] The comprehensive risk index calculation module is interconnected with the S5 dynamic threshold model. When the threshold model detects that the abnormal state duration exceeds the set window, the risk index will automatically increase according to the abnormal persistence. At the same time, if the wind speed increases or the ventilation mode is switched to the strong exhaust mode, the diffusion risk weight is automatically reduced to achieve dynamic balance of risk.

[0133] The comprehensive risk index is divided into three levels according to the interval value: (Note: This numerical division comes from long-term operation data statistics and experimental calibration results, not subjective setting);

[0134] Warning level: low risk, Ra≤0.4; When the risk level is the warning level: execute local ventilation control, start the auxiliary exhaust device in the cabin, and upload the risk prompt information to the operation monitoring center. This stage is early prevention, mainly to prevent gas accumulation.

[0135] Alert level: medium risk, 0.4

[0136] Emergency level: high risk, Ra>0.7; When the risk level is the emergency level: automatically execute the energy storage cabin emergency linkage control action, including: automatically cut off the power supply of the energy storage cabin to eliminate the risk of electrical fire; Start the emergency ventilation and fire extinguishing system; Trigger the linkage alarm system, send an emergency alarm to the control center and safety system synchronously.

[0137] The hierarchical strategy is triggered based on the continuous change of risk level, without discrete jumps, avoiding the problem of false triggering in boundary states. Each control action is derived from the safety protection standards and experimental verification process of the energy storage system, and has practical engineering basis. While executing the hierarchical control, the comprehensive risk index and hierarchical response are recorded to the database, including the time stamp, risk distribution map and control action feedback information.

[0138] Step S7: Based on the historical distribution of concentration anomalies and leakage anomalies, the monitoring coverage is evaluated, and the layout of the monitoring points is updated.

[0139] The updating of the layout of the monitoring points includes: collecting and storing the historical distribution data of concentration anomalies and leakage anomalies, statistically analyzing the spatial coordinates and occurrence frequency, mapping the historical distribution data to the monitoring area defined by the geometric structure model of the energy storage cabin, calculating the monitoring coverage index according to the spatial density of the corresponding anomalies in the historical distribution data and the spatial distance from the monitoring points to the anomaly positions; identifying the areas with insufficient monitoring coverage, determining the position adjustment amount of the monitoring points based on spatial reconstruction and reachability analysis, generating coordinate suggestions for the migration or addition of monitoring points, and updating the layout configuration of the monitoring points.

[0140] In an embodiment of the present application, step S7 dynamically evaluates the spatial coverage degree of the current monitoring system for potential leakage events by statistically analyzing the historical monitoring data during the operation of the energy storage cabin, and optimizes the layout of the monitoring points accordingly, realizing the self-evolution and updating of the long-term monitoring capability. The layout of the monitoring points is dynamically optimized according to the actual risk distribution, so that the monitoring system maintains long-term consistency in spatial balance and risk response capability.

[0141] The system statistically analyzes the historical data by type: first-level classification, concentration anomaly, leakage anomaly; second-level classification, anomaly level (mild, moderate, severe) and duration; third-level classification, corresponding airflow state and operating condition (ventilation mode, load level, etc.).

[0142] Map the historical anomaly data to the geometric structure model of the energy storage cabin according to the spatial coordinates, and form the anomaly event distribution map. Each event is represented by its central coordinate as a point in the three-dimensional model, with a weight density determined by the anomaly level and occurrence frequency.

[0143] Monitoring coverage is used to quantify the spatial response capability of the current monitoring point layout to anomaly events. It is defined as: monitoring coverage = (volume of anomaly area that can be effectively detected by the monitoring points) / (total volume of historical anomaly area); it is obtained by calculating the overlap degree of the monitoring point detection radius (based on the detectable distance of the airflow vector model) and the spatial distribution of anomaly events. The detection radius of each monitoring point is determined by the sensor sensitivity, ventilation state and airflow direction, which is not artificially set.

[0144] Quantitative calculation process: According to the energy storage cabin geometric model, the spatial distribution matrix of the monitoring points is established; using the historical abnormal point coordinate set, a three-dimensional abnormal density field is generated; the coverable volume of each monitoring point to the abnormal density field is calculated; the coverable volume of all monitoring points is integrated to form the overall monitoring coverage index. When the coverage is lower than the system set optimization threshold (for example, 80%, which is determined by historical statistics), the layout updating process is triggered.

[0145] The spatial clustering analysis is performed on the area with insufficient monitoring coverage to identify the area with high abnormal density and low monitoring sensitivity.

[0146] The typical insufficient coverage area includes: airflow dead angle (such as the opposite side of the air inlet or the structure shielding area); heat accumulation area (battery pack gap, cabin bottom); uneven flow disturbance area (local vortex formation). Based on the cabin structure and ventilation model, spatial reconstruction analysis is performed in the identified insufficient coverage area to determine the flow path of the gas in the area and the accessibility of the sensor. The shortest path algorithm is used to calculate the accessibility index of the monitoring point to determine whether the area can be covered by migrating existing monitoring points or new sensor nodes need to be added.

[0147] According to the spatial reconstruction result, the coordinate suggestion for monitoring point migration or addition is generated: if the coverage can be optimized by adjusting the position, the migration coordinate suggestion is output; if new monitoring points need to be added, the addition coordinate suggestion is output; each suggestion includes specific coordinates, expected coverage improvement value and implementation priority. The update result is written into the monitoring point layout configuration file as the initial input for the next period layout optimization, forming a long-term self-evolution mechanism.

[0148] The monitoring point layout update is completely generated based on the measured abnormal data and the airflow simulation result, without using artificial experience or subjective set adjustment rules. All migration or addition suggestions have clear spatial basis and quantifiable improvement effect. The effect of each layout update can be verified by recalculating the monitoring coverage index, and if the coverage improvement rate reaches the expected value, the update scheme is confirmed to be effective.

[0149] Embodiment 2, a hydrogen sulfide leakage anomaly monitoring system, as shown in Figure 1 The system includes the following modules:

[0150] Data monitoring module: real-time acquisition of data of each monitoring point of the energy storage cabin of the energy storage power station, including environmental parameters, ventilation equipment state, concentrations of hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride, and synchronous acquisition of health state parameters of the battery;

[0151] Airflow correction module: obtain airflow parameters according to environmental parameters and ventilation equipment state, obtain the geometric structure model of the energy storage cabin, select the airflow vector model, calculate the gas diffusion disturbance coefficient, correct the hydrogen sulfide concentration based on the airflow field, and obtain the corrected concentration;

[0152] Twin inversion module: based on the environment parameters and health state parameters, the battery thermal distribution is obtained, combined with the revised concentration, air flow parameters and geometric structure model, a digital twin model of double physical field is established, and the spatial position of concentration anomaly, the diffusion path and concentration distribution of hydrogen sulfide at the preset time node are determined through inversion calculation;

[0153] Correlation judgment module: based on the revised concentration, and the concentration of hydrogen, carbon monoxide and hydrogen fluoride, the corresponding concentration change rate curve is constructed, and the time sequence correlation and proportional relationship of the hydrogen sulfide concentration change rate curve and the rest of the concentration change rate curve are analyzed, based on the battery fault mode library, the fault type of the concentration anomaly is judged;

[0154] Leakage confirmation module: based on the health state parameters, environment parameters and fault type, the dynamic threshold model is adjusted, the revised concentration and hydrogen sulfide concentration change rate curve are monitored through the dynamic threshold model, and the leakage anomaly is determined;

[0155] Linkage control module: according to the health state parameters, the corresponding grade of the revised concentration, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution and the fault type, the comprehensive risk index is calculated, and the grading linkage control strategy is executed based on the comprehensive risk index;

[0156] Monitoring update module: based on the historical distribution of concentration anomaly and leakage anomaly, the monitoring coverage is evaluated, and the position of the monitoring point is updated.

[0157] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A hydrogen sulfide leak anomaly monitoring method characterized by, The method comprises the following steps: Real-time acquisition of data of each monitoring point of the energy storage cabin of the energy storage power station, including environmental parameters, ventilation equipment state, hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride concentration, and synchronous acquisition of battery health state parameters; According to the environmental parameters and the ventilation equipment state, the air flow parameters are obtained, the geometric structure model of the energy storage cabin is obtained, and the air flow vector model is selected, the gas diffusion disturbance coefficient is calculated, the hydrogen sulfide concentration is corrected by the air flow field, and the corrected concentration is obtained; Based on the environmental parameters and the health state parameters, the battery heat distribution is obtained, the corrected concentration, the air flow parameters and the geometric structure model are combined to establish a digital twin model of double physical fields, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution of hydrogen sulfide at the preset time node are determined through inversion calculation; Based on the corrected concentration, the concentration change rate curves of hydrogen, carbon monoxide and hydrogen fluoride are constructed, the time sequence correlation and proportional relationship between the hydrogen sulfide concentration change rate curve and the rest concentration change rate curves are analyzed, and the fault type of the concentration anomaly is determined based on the battery fault mode library; Based on the health state parameters, the environmental parameters and the fault type, the dynamic threshold model is adjusted, the corrected concentration and the hydrogen sulfide concentration change rate curve are monitored through the dynamic threshold model, and the leakage anomaly is determined; According to the health state parameters, the corresponding grade of the corrected concentration, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution and the fault type, the comprehensive risk index is calculated, and the grading linkage control strategy is executed based on the comprehensive risk index; Based on the historical distribution of the concentration anomaly and the leakage anomaly, the monitoring coverage is evaluated, and the layout of the monitoring points is updated.

2. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The calculation of the gas diffusion disturbance coefficient comprises: the air flow parameters are composed of the monitoring point wind speed vector, the air volume distribution and the ventilation equipment state; according to the geometric structure model of the energy storage cabin and the typical operation condition, the computational fluid dynamics method is used to construct the basic air flow vector model of the energy storage cabin offline, and the flow field feature database under different ventilation equipment states is generated; the current air flow parameters are compared with the flow field feature database, the corresponding air flow vector model is selected in the basic air flow vector model, the turbulent kinetic energy distribution and the local vector of each region in the energy storage cabin are obtained through the air flow vector model, the local vector is differentially calculated with the measured wind speed vector obtained by the monitoring point, and the air flow deviation field is formed; based on the air flow deviation field and the distribution of the turbulent kinetic energy, the least square fitting algorithm is used to calculate the gas diffusion disturbance coefficient.

3. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The gas flow field correction of the hydrogen sulfide concentration comprises: the hydrogen sulfide concentration measured by the monitoring point is spatially corresponding to the air flow vector model, the gas diffusion disturbance coefficient of the monitoring point is obtained; based on the gas diffusion disturbance coefficient and the wind speed vector, the concentration drift deviation of the hydrogen sulfide concentration measurement caused by the air flow is calculated; the measured hydrogen sulfide concentration is additively compensated and corrected through the concentration drift deviation, and the corrected concentration is obtained.

4. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The establishing of the digital twin model of the double physical fields comprises: obtaining a battery heat distribution through temperature and health state parameters in the environmental data, the health state parameters comprising a charge-discharge rate, a voltage, voltage consistency, internal resistance, and historical charge-discharge cycle times; coupling the battery heat distribution, airflow parameters, and a geometric structure model to construct a double physical field coupling domain inside the energy storage cabin; in the double physical field coupling domain, a coupling model of a heat field equation, a flow field equation, and a concentration equation is established by using a finite volume method, and a battery module heat dissipation channel and a ventilation air outlet are introduced as boundary conditions; a modified concentration is introduced into the coupling model as a passive scalar to realize synchronous evolution simulation of the heat field, the flow field, and the concentration field, and form the digital twin model of the double physical fields.

5. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The inversion calculation comprises: inputting environmental parameters and health state parameters in the digital twin model, numerically solving through the coupling model to obtain a distribution of airflow vectors and a predicted concentration of hydrogen sulfide inside the energy storage cabin; performing difference calculation on the predicted concentration of hydrogen sulfide and the modified concentration, determining the spatial position of the concentration anomaly based on the spatial residual distribution of the two, and determining a concentration anomaly spatial position by using a least mean square error inversion algorithm; calculating a gas transport velocity field according to the distribution of the airflow vectors and the turbulent kinetic energy distribution; solving a non-steady-state concentration transport equation in the transport velocity field to obtain a concentration distribution of hydrogen sulfide over time; and generating a diffusion path of hydrogen sulfide in a preset time window by spatially fitting concentration isosurfaces at different times.

6. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The determining of the fault type of the concentration anomaly comprises: extracting feature parameters including peak occurrence time, rising rate, and change duration based on each concentration change rate curve; obtaining a joint feature vector of multiple gases by analyzing the time sequence correlation and concentration proportion relationship between the hydrogen sulfide concentration change rate curve and the hydrogen, carbon monoxide, and hydrogen fluoride concentration change rate curves; and matching the joint feature vector with a pre-established battery fault mode library to determine the fault type.

7. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The adjusting of the dynamic threshold model comprises: obtaining a battery health state deviation, an environmental disturbance coefficient, and a fault risk weight based on the health state parameters, the environmental parameters, and the fault type of the battery respectively, performing normalized weighted calculation on the battery health state deviation, the environmental disturbance coefficient, and the fault risk weight to obtain a threshold adjustment factor, and correcting the concentration threshold and the change rate threshold according to the threshold adjustment factor; and the dynamic threshold model is updated adaptively by a time weighting and historical data backtracking algorithm.

8. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The executing of the hierarchical linkage control strategy comprises: dividing a hydrogen sulfide leakage event into a pre-warning level, a warning level, and an emergency level according to a risk level corresponding to a comprehensive risk index; executing local ventilation control and uploading risk prompt information when the risk level is the pre-warning level; starting an independent exhaust device of a leakage area and notifying an operation and maintenance system to enter a monitoring intensive mode when the risk level is the warning level; and executing energy storage cabin emergency linkage control actions including power cutoff and ventilation fire extinguishing startup when the risk level is the emergency level; and the comprehensive risk index is formed by weighting a concentration weight, an electrical risk weight, and a diffusion risk weight.

9. The hydrogen sulfide leak anomaly monitoring method of claim 1, wherein, The layout of the update monitoring point comprises: collecting and storing historical distribution data of concentration anomalies and leakage anomalies, counting according to spatial coordinates and occurrence frequency, mapping the historical distribution data to the monitoring area defined by the energy storage cabin geometric structure model, and calculating the monitoring coverage index according to the spatial density of the corresponding anomaly of the historical distribution data and the spatial distance from the monitoring point to the anomaly position; identifying the area with insufficient monitoring coverage, determining the position adjustment amount of the monitoring point based on spatial reconstruction and accessibility analysis, generating coordinate suggestions for monitoring point migration or addition, and updating the monitoring point layout configuration.

10. A hydrogen sulfide leak anomaly monitoring system, comprising: The system applies the hydrogen sulfide leakage anomaly monitoring method of any one of claims 1 to 9, comprising: A data monitoring module: real-time collection of data of each energy storage cabin monitoring point of the energy storage power station, including environmental parameters, ventilation equipment state, hydrogen sulfide, hydrogen, carbon monoxide and hydrogen fluoride concentration, and synchronous collection of battery health state parameters; An airflow correction module: obtaining airflow parameters according to environmental parameters and ventilation equipment state, obtaining the geometric structure model of the energy storage cabin, selecting an airflow vector model, calculating a gas diffusion disturbance coefficient, correcting the hydrogen sulfide concentration in the airflow field, and obtaining a corrected concentration; A twin inversion module: obtaining battery heat distribution based on environmental parameters and health state parameters, combining corrected concentration, airflow parameters and geometric structure model to establish a digital twin model of double physical fields, and determining the spatial position of concentration anomalies, as well as the diffusion path and concentration distribution of hydrogen sulfide at a preset time node through inversion calculation; An association judgment module: based on the corrected concentration, and the concentration of hydrogen, carbon monoxide and hydrogen fluoride, a corresponding concentration change rate curve is constructed, and the time sequence association and proportional relationship of the hydrogen sulfide concentration change rate curve and the remaining concentration change rate curves are analyzed, and based on the battery fault mode library, the fault type of the concentration anomaly is judged; A leakage confirmation module: based on the health state parameters, environmental parameters and fault type, adjusting the dynamic threshold model, monitoring the corrected concentration and hydrogen sulfide concentration change rate curve through the dynamic threshold model, and determining the leakage anomaly; A linkage control module: according to the health state parameters, the corresponding level of the corrected concentration, the spatial position of the concentration anomaly, the diffusion path and the concentration distribution, and the fault type, a comprehensive risk index is calculated, and a hierarchical linkage control strategy is executed based on the comprehensive risk index; A monitoring update module: based on the historical distribution of concentration anomalies and leakage anomalies, the monitoring coverage is evaluated, and the position of the monitoring point is updated.

Images