CN-121997763-A - Hydrogen elimination efficiency testing system and method based on multi-sensor collaborative monitoring

CN121997763ACN 121997763 ACN121997763 ACN 121997763ACN-121997763-A

Abstract

The invention discloses a system and a method for testing the hydrogen elimination efficiency based on multi-sensor collaborative monitoring, wherein the method comprises the steps of obtaining the operation parameters of a hydrogen elimination means; and taking the operation parameters of the dehydrogenation means as the input of a pre-trained dehydrogenation efficiency prediction model, and outputting the corresponding dehydrogenation efficiency of the dehydrogenation means. The method effectively solves the problems of large coupling interference, poor data synchronism and insufficient prediction capability in the traditional test, and remarkably improves the accuracy, robustness and intelligent level of the hydrogen elimination efficiency evaluation.

Inventors

LI JINGFA
DENG XINKE
YANG ZIKAI
HUANG HUIJIE
FAN CHAOYANG
WANG YANXIN
XIAO JIALE
YU BO

Assignees

长江大学

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. The method for testing the hydrogen elimination efficiency based on the multi-sensor collaborative monitoring is characterized by comprising the following steps of: Obtaining the operation parameters of the hydrogen elimination means; and taking the operation parameters of the dehydrogenation means as the input of a pre-trained dehydrogenation efficiency prediction model, and outputting the corresponding dehydrogenation efficiency of the dehydrogenation means.
2. The method of claim 1, wherein the means for hydrogen elimination comprises at least one of mechanical ventilation, inerting, and catalytic oxidation.
3. The method of claim 2, wherein the operating parameters of the hydrogen abatement means include mechanical ventilation parameters, inerting parameters, and catalytic oxidation parameters; wherein, the mechanical ventilation parameters comprise ventilation air quantity, ventilation direction and ventilation mode; the inerting treatment parameters include inert gas flow, inert gas type and inert gas injection mode; catalytic oxidation parameters include catalyst temperature, catalyst type, and catalyst loading.
4. A method according to claim 3, wherein the hydrogen-elimination efficiency prediction model comprises: The feature grouping embedding layer is used for grouping according to the business and physical significance of the operation parameters, compressing the operation parameters of each group into a low-dimensional dense feature vector, and respectively obtaining three independent feature vectors of a ventilation vector, an inerting vector and a catalysis vector; The multi-head cross attention module is used for capturing and quantifying the mutual influence among the feature vectors corresponding to different hydrogen elimination means and generating interference perception fusion features; the interference sensing gating layer is used for analyzing the interference patterns contained in the interference sensing gating layer according to the interference sensing fusion characteristics, outputting a corresponding interference index for the identified interference patterns, and correcting a predicted result for the hydrogen elimination efficiency according to the interference index; The first interference mode is that the ventilation air quantity is higher than a set value to blow off inert gas; the second interference mode is that the flow rate of the inert gas is higher than a set value, and the temperature of the catalyst does not reach the set value; the third interference mode is that the ventilation air quantity is higher than a set value, and the temperature of the catalyst does not reach the set value; Interference mode four, interference mode one, interference mode two, interference mode three exist at the same time.
5. The method of claim 4, wherein capturing and quantifying interactions between feature vectors corresponding to different hydrogen elimination means by the multi-headed cross-attention module comprises: Taking the ventilation vector as a query vector, taking the inerting vector or the catalysis vector as a key vector, taking the inerting vector or the catalysis vector as a value vector, calculating the similarity between the query vector and the key vector, generating an attention weight matrix, carrying out weighted summation on the value vector by utilizing the attention weight matrix, and outputting a first fusion feature vector; taking the inerting vector as an inquiry vector, taking the ventilation vector or the catalysis vector as a key vector, and taking the ventilation vector or the catalysis vector as a value vector to obtain a second fusion characteristic vector by a method of obtaining the first fusion characteristic vector; taking the catalytic vector as an inquiry vector, taking the ventilation vector or the inerting vector as a key vector, and taking the ventilation vector or the inerting vector as a value vector to obtain a third fusion feature vector by a method of obtaining a first fusion feature vector; And splicing the first fusion feature vector, the second fusion feature vector and the third fusion feature vector to obtain the interference perception fusion feature.
6. The method of claim 5, wherein the specific process of training the hydrogen-elimination efficiency prediction model comprises the steps of: S1, simulating a leakage scene of hydrogen or hydrogen-doped mixed gas in a region to be tested, and providing stable and repeatable leakage working conditions; S2, executing a dehydrogenation means in the region to be detected, and recording operation parameters corresponding to the dehydrogenation means; S3, monitoring the hydrogen concentration change rates of different monitoring points in the to-be-detected area in real time, and collecting the hydrogen concentration and the environmental parameters of the corresponding monitoring points when the hydrogen concentration change rates are smaller than a preset stability threshold value; S4, decoupling correction is carried out on the hydrogen concentration measured value by using an environment interference correction model based on a plurality of groups of sensor sampling data, and smoothing filtering processing is carried out on the corrected data to obtain corrected and filtered data; s5, calculating the instantaneous hydrogen elimination efficiency at each moment in a preset evaluation duration based on the corrected and filtered data by using an instantaneous hydrogen elimination efficiency dynamic evolution model, and constructing a hydrogen elimination efficiency curve; S6, identifying an efficient region, a saturated attenuation region and a dead zone in a hydrogen elimination efficiency curve, extracting instantaneous hydrogen elimination efficiency in the efficient region, and calculating an integral average value of the instantaneous hydrogen elimination efficiency in the efficient region; s7, repeating the steps S1-S6 for a plurality of times, and calculating arithmetic mean of a plurality of integral mean values to obtain real hydrogen elimination efficiency; s8, inputting the operation parameters corresponding to the dehydrogenation means into a dehydrogenation efficiency prediction model to obtain predicted dehydrogenation efficiency; S9, training a hydrogen elimination efficiency prediction model by using a loss function containing an interference penalty term based on predicted hydrogen elimination efficiency until the loss function is not reduced; The loss function containing the interference penalty term is specifically: in the formula, Representing the total loss function of the device, Indicating the predicted error in the hydrogen-elimination efficiency, Representing the hydrogen-elimination efficiency prediction error weight coefficient, Indicating the prediction error of the degree of interference, Representing the interference degree prediction error weight coefficient; Represents the number of hydrogen elimination means samples during training, The index of the sample is represented and, Represent the first The actual hydrogen-elimination efficiency corresponding to the individual samples, Represent the first Predicted hydrogen elimination efficiency corresponding to each sample; Represent the first The actual interference index corresponding to the individual samples, Represent the first The predicted interference index corresponding to each sample, Is a penalty coefficient.
7. The method of claim 6, wherein the environmental parameters include temperature, pressure, humidity, and wind speed.
8. The method of claim 7, wherein the environmental interference correction model is specifically: in the formula, In order to correct the concentration of hydrogen after filtration, As a measure of the concentration of hydrogen gas, , , , The temperature, pressure, humidity and wind speed are respectively collected in real time; , , the parameters are standard working condition parameters corresponding to temperature, pressure and humidity; , , Is a first order term coefficient; , Is a coupling term coefficient.
9. The method according to claim 6, wherein the instantaneous hydrogen elimination efficiency dynamic evolution model is specifically: in the formula, Is that The instant hydrogen-eliminating efficiency at the moment, Is that The corrected and filtered hydrogen concentration at the moment in time, Representing the operating mode normalization factor.
10. A system based on the multi-sensor collaborative monitoring-based hydrogen desorption efficiency testing method according to any one of claims 1-9, comprising: The data acquisition module is used for acquiring the operation parameters of the hydrogen canceling means; and the efficiency prediction module is used for taking the operation parameters of the dehydrogenation means as the input of a pre-trained dehydrogenation efficiency prediction model and outputting the corresponding dehydrogenation efficiency of the dehydrogenation means.

Description

Hydrogen elimination efficiency testing system and method based on multi-sensor collaborative monitoring Technical Field The invention belongs to the technical field of hydrogen energy safety and risk prevention and control, and particularly relates to a hydrogen elimination efficiency test system and method based on multi-sensor collaborative monitoring. Background With the wide application of the hydrogen energy industry in the fields of energy, traffic and industry, the safety guarantee technology of the hydrogen energy industry is increasingly paid attention to. Because hydrogen has the physical characteristics of wide explosion limit, low ignition energy, high diffusion speed and the like, once leakage occurs in a limited space or a semi-open environment, the hydrogen is easy to accumulate to form explosive mixed gas, and the hydrogen forms serious threat to personnel safety and equipment and facilities. Therefore, the adoption of efficient hydrogen elimination measures aiming at leakage scenes is a key link for reducing the application risk of hydrogen energy and guaranteeing the operation safety. At present, the dehydrogenation means commonly adopted in engineering practice mainly comprise mechanical ventilation, inerting treatment, catalytic oxidation and the like. Mechanical ventilation dilutes and discharges hydrogen by forced gas flow, inerting reduces oxygen concentration with inert gas to inhibit combustion, and catalytic oxidation converts hydrogen to water with a catalyst. In order to evaluate the actual effectiveness of these means, the existing technical solutions generally rely on specific experimental cabins or test platforms, and the concentration change in the hydrogen elimination process is monitored by simulating the leakage working condition, so as to be used as the basis for evaluating the performance of the hydrogen elimination equipment or making a safety strategy. However, the conventional hydrogen elimination efficiency test system still has a plurality of key technical bottlenecks in practical application, and is difficult to meet the requirements of high-precision and intelligent evaluation. First, existing systems lack multi-dimensional decoupling correction capabilities for complex environmental factors. The traditional monitoring mode mostly relies on single linear temperature compensation, and cross coupling interference among environmental factors is ignored. Particularly, under the working conditions of mechanical ventilation or large flow leakage, the change of wind speed (flow field) in a region to be tested can obviously change the thermal balance state of the sensor surface, and serious nonlinear measurement errors can be generated when high-temperature and high-humidity environments are superposed, so that the accuracy of test data is greatly reduced. Secondly, the synchronism and reliability of the existing data acquisition scheme are insufficient. On a hardware architecture, the traditional scheme relies on single-point sensor sampling, lacks a consistency check mechanism based on hardware redundancy, and once the sensor is subjected to 'poisoning', aging drift or sudden faults, the system cannot identify abnormal data. In the aspect of data transmission, a software polling mode is adopted to read data of each channel, so that millisecond-level time deviation exists among different sensors, microsecond-level strict synchronization is difficult to realize, and when the communication bandwidth fluctuates, an effective local buffer (FIFO) and breakpoint continuous transmission mechanism are lacked, so that the loss of key transient sequence data is extremely easy to cause, and the capture of a hydrogen elimination dynamic process is seriously influenced. Finally, existing methods lack intelligent prediction means based on historical data. The traditional testing method only stays at the 'post calculation' level of the current experimental data, a high-dimensional mathematical model is not established by utilizing the accumulated historical data, and a deep mapping relation among leakage parameters, environmental conditions and hydrogen elimination efficiency cannot be mined. This results in a system that does not have the ability to predict and optimize the unknown conditions in the future online, often requiring extensive repeated physical experiments to verify the scheme, which is inefficient and costly. Aiming at the problems, development of a hydrogen elimination efficiency testing system and method with multidimensional environment decoupling correction, hardware-level high-precision synchronization and AI prediction capability is urgently needed to improve the accuracy, robustness and intelligence level of hydrogen energy safety test. Disclosure of Invention Aiming at the defects in the prior art, the system and the method for testing the hydrogen elimination efficiency based on the multi-sensor collaborative monitoring solve the problems of testing deviation c