CN-122016976-A - Collaborative calibration method and system for electrochemical gas sensor

CN122016976ACN 122016976 ACN122016976 ACN 122016976ACN-122016976-A

Abstract

The invention relates to a collaborative calibration method of an electrochemical gas sensor, which comprises the following steps of synchronously collecting electrochemical response signals and environmental parameters of a target sensor under preset environmental conditions to construct various feature combinations, wherein the response signals are obtained under at least two hardware collection configurations, training correction models of various architectures for each feature combination, quantifying the calculation resource cost of single prediction of each feature combination, constructing a multidimensional decision space representing the mapping relation of model types, feature combinations, hardware configuration, precision and calculation cost by combining corresponding hardware configuration information after evaluating the calibration precision of each model, and matching the optimal correction models, feature combinations and hardware configuration from the space according to the requirements of specific application scenes on the calculation cost and the precision to realize the accurate and efficient calibration of gas concentration. Through the design, the invention realizes systematic balance between calibration precision and calculation resource consumption.

Inventors

BAI XIANG
ZHANG JIAHUI
LI HUAJIAN
CHEN YIJIE
ZHANG XUAN
XU YINLIANG

Assignees

山西省能源互联网研究院
清华大学深圳国际研究生院

Dates

Publication Date: 20260512
Application Date: 20260303

Claims (10)

1. The collaborative calibration method of the electrochemical gas sensor is characterized by comprising the following steps of: S1, selecting a target sensor; S2, synchronously acquiring an electrochemical response signal and at least one environmental parameter of the target sensor under a preset environmental condition to construct a plurality of characteristic combinations, wherein the electrochemical response signal is acquired under at least two different hardware acquisition configurations; S3, respectively training correction models with different architectures according to each characteristic combination, and quantifying the calculation resource cost required by each trained correction model when calibration prediction is carried out once; S4, evaluating the calibration precision of each trained correction model under the corresponding characteristic combination by using a test set; s5, based on the calculation resource cost and the calibration precision of each correction model, combining hardware acquisition configuration information corresponding to each feature combination to construct a multi-dimensional decision space, wherein the decision space is used for representing the mapping relation among the types of the correction models, the feature combinations, the hardware acquisition configuration, the calibration precision and the calculation resource cost; And S6, matching the optimal correction model, feature combination and hardware acquisition configuration from the decision space according to the requirements of the application scene on the calculation resource cost and the calibration precision, and performing gas concentration calibration on the target sensor.
2. The co-calibration method according to claim 1, wherein the target sensor comprises a two-electrode electrochemical gas sensor, a three-electrode electrochemical gas sensor, and a four-electrode electrochemical gas sensor.
3. The collaborative calibration method according to claim 1, wherein the temperature interval of the predetermined environmental conditions is-30 ℃ to 30 ℃.
4. The co-calibration method according to claim 1, wherein the "at least one environmental parameter" comprises an ambient temperature and an ambient humidity.
5. The co-calibration method according to claim 1, wherein the acquiring under at least two different hardware acquisition configurations comprises sampling the electrochemical response signal with at least two different resolutions by an analog-to-digital converter.
6. The collaborative calibration method according to claim 1, wherein the "correction models for multiple different architectures" include at least two of a support vector regression model, a multi-layer perceptron model, and a convolutional neural network model.
7. The collaborative calibration method according to claim 1, wherein the "computational resource costs" are quantified by the number of floating point operations required to perform a calibration prediction.
8. The collaborative calibration method according to claim 1, wherein the step S6 further comprises, for the edge deployment scenario with limited computing resources, preferentially selecting a correction model that meets a preset minimum calibration accuracy threshold and has minimum computing resource cost in the decision space, and simultaneously selecting a feature combination and a hardware acquisition configuration corresponding to the correction model.
9. A co-calibration system for an electrochemical gas sensor, comprising: the signal acquisition module is used for acquiring electrochemical response signals of the target sensor under at least two hardware acquisition configurations; the environment parameter acquisition module is used for acquiring at least one environment parameter; the memory is used for storing a multi-dimensional decision space which is generated based on historical data training collected under different hardware collection configurations and represents the mapping relation among the correction model types, the feature combinations, the hardware collection configurations, the calibration precision and the calculation resource cost; The processor is coupled to the signal acquisition module, the environment parameter acquisition module and the memory, and is configured to construct a feature combination according to the electrochemical response signal and the environment parameter which are acquired currently, match an optimal correction model, the feature combination and the hardware acquisition configuration from the decision space according to the requirements of an application scene on the calculation resource cost and the calibration precision, process the data acquired currently according to the feature combination and the hardware acquisition configuration corresponding to the optimal correction model and output a gas concentration calibration value.
10. The co-calibration system according to claim 9, wherein the signal acquisition module comprises an analog-to-digital converter, the "acquisition in at least two hardware acquisition configurations" specifically being configured to sample the electrochemical response signal at least two different resolutions.

Description

Collaborative calibration method and system for electrochemical gas sensor Technical Field The invention relates to the technical field of electrochemical sensors, in particular to a collaborative calibration method and a collaborative calibration system of an electrochemical gas sensor. Background In low-temperature environments such as polar regions and mountains, concentration monitoring of key gases such as oxygen directly relates to human health and operation safety, for example, accurate oxygen data is provided for polar region scientific investigation team members and high-altitude mountain climbers so as to prevent hypoxia accidents. Electrochemical gas sensors are widely used for their low cost, but in sub-zero environments their performance is significantly reduced, particularly as a result of temperature-induced signal drift. This phenomenon is particularly pronounced in a two-electrode electrochemical gas sensor that is simpler in construction and less costly. In particular, a two-electrode electrochemical oxygen sensor relies on the principle of electrochemical reaction, and is typically composed of a working electrode, a reference electrode and an electrolyte. Under normal temperature (20-25 ℃), oxygen molecules undergo reduction reaction on the surface of a working electrode to generate electrochemical response signals (usually voltage) in direct proportion to the oxygen concentration, and the oxygen concentration can be quantitatively measured after the electrochemical response signals are processed by a signal conditioning circuit. However, in low temperature environments (especially below 0 ℃) significant decay in performance of such sensors occurs, manifested by reduced response sensitivity, prolonged response time, baseline drift, and non-linear enhancement. The basic reasons are that the electrolyte has various physical and chemical changes inside the sensor caused by low temperature (1) the electrolyte has the characteristics of increased viscosity, reduced ionic conductivity and even phase change (such as crystallization or gelation) to prevent ion transmission, (2) the electrochemical reaction kinetics is slowed down, the reaction rate constant is exponentially reduced according to an Arrhenius equation, the oxygen reduction reaction rate is slowed down, the response speed and sensitivity are reduced, (3) the gas diffusion is limited, the gas molecular diffusion coefficient is reduced, the transmission rate of oxygen to the electrode surface is reduced, the response time is further prolonged, and (4) the state of the electrode surface is changed, the adsorption layer structure is changed, the catalytic activity is reduced or the water vapor condensation is reduced, so that the normal reaction is influenced. Under the multi-factor coupling effect, the double-electrode electrochemical oxygen sensor is difficult to maintain stable and reliable measurement performance in extremely low-temperature environments (such as aerospace, plateau scientific investigation and cold region industry), cannot directly use normal-temperature calibration parameters for concentration calculation, and must be corrected through temperature compensation or intelligent calibration technology. Some of the prior art studies have attempted to employ three-electrode or four-electrode systems with auxiliary electrodes, or complex machine learning algorithms (e.g., convolutional neural networks, deep artificial neural networks) to compensate for temperature effects. However, the existing methods have obvious limitations that either the hardware-level auxiliary electrode is relied on to provide additional information, but a complex driving circuit is needed and the cost is high, or the computing-intensive deep learning model is relied on, so that the method is difficult to deploy in real time in the edge equipment with limited resources. The prior art can not effectively solve the dual constraint between the limit temperature and the limited calculation force, and can not simultaneously consider the model reasoning speed and the sensing precision. Therefore, a hardware algorithm collaborative calibration scheme capable of automatically adjusting algorithm complexity according to hardware resolution, and dynamically balancing accuracy and efficiency is needed. In addition, the current calibration method faces two major core challenges in practical deployment: the contradiction between the complexity of the model and the computing capacity of the edge, namely, although an advanced calibration model (such as a deep learning network) has high precision potential, the huge computing demands (such as floating point operand and memory occupation) and the limited processing and storage resources of the edge equipment (especially a low-power microcontroller) have sharp contradictions, so that the real-time operation in a portable or remote sensing system with quick response and low energy consumption is difficult. The data p