A method and system for calibrating harmful gas concentration

Abstract

The present application relates to the technical field of gas sensor calibration, and specifically discloses a harmful gas concentration calibration method and system, the method comprising: constructing a time domain drift feature channel and a component cross interference feature channel according to a target harmful gas original time sequence monitoring data set, to obtain a double-channel state representation; based on a preset historical steady-state baseline data, performing error traceability separation on the double-channel state representation to independently distinguish the current measurement deviation into a zero point drift component, a sensitivity drift component and a cross interference component; establishing a dynamic coupling calibration model to output a calibrated harmful gas concentration value; continuously evaluating the current compensation confidence level according to the intermediate parameters of the calibration process, and triggering a standard gas review prompt when the compensation confidence level is lower than a preset threshold; the present application can improve the detection sensitivity of the gas sensor in a multi-harmful gas coexistence scene, and reduce the false alarm and missed alarm risks.

Description

Technical Field

[0001] This invention relates to the field of gas sensor calibration technology, and in particular to a method and system for calibrating the concentration of harmful gases. Background Technology

[0002] Hazardous gas concentration monitoring plays a crucial role in industrial safety, environmental monitoring, coal mining, and petrochemical industries, serving as a core technological means to ensure personnel safety and prevent major accidents. As a key component in concentration detection, the long-term measurement accuracy of gas sensors directly impacts the reliability and early warning effectiveness of monitoring systems. Currently, maintaining the accuracy of on-site hazardous gas concentration measurements primarily employs two methods: one is periodic calibration using standard gas cylinders according to national metrological regulations; the other is static reading compensation using environmental parameters. Existing solutions typically combine these two approaches. However, this particular solution integrates environmental parameter acquisition modules such as temperature, humidity, and air pressure into the gas sensing system. It utilizes pre-calibrated compensation coefficients to perform real-time static correction of sensor readings. Simultaneously, standard gas is introduced at regular intervals, and the sensor response value is compared with the standard gas concentration. Based on this comparison, sensor parameters are adjusted to eliminate accumulated deviations.

[0003] However, this existing technical solution has a core flaw: it cannot independently identify and decouple measurement deviations from three different physical sources: sensor aging drift, multi-gas cross-sensitivity, and environmental temperature and pressure fluctuations. During long-term sensor operation, the natural aging of the sensitive material leads to baseline drift and sensitivity decay. Coexisting components in the measured mixed gas exhibit cross-sensitivity responses to the target sensor. Environmental temperature and pressure fluctuations cause instantaneous changes in the sensor's response characteristics. These three types of errors exhibit different patterns of change and have different compensation directions. The existing solution treats these errors as a single, unified correction, which means that compensation measures for one type of error may exacerbate another. Calibration accuracy deteriorates rapidly over time, and long-term measurement errors accumulate, easily leading to false alarms or missed alarms. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a method and system for calibrating the concentration of harmful gases, so as to solve the problem that in the existing harmful gas concentration calibration scheme, the measurement deviations from three different physical sources, namely sensor aging and drift, multi-gas cross-sensitivity and environmental temperature and pressure fluctuations, are treated together and cannot be independently identified and decoupled, resulting in the rapid deterioration of calibration accuracy over time and the continuous accumulation of long-term measurement errors.

[0005] This invention provides a method for calibrating the concentration of harmful gases, comprising:

[0006] Step 1: Acquire multi-channel raw sensing signals of the target harmful gas through a gas sensing array, and simultaneously collect real-time environmental parameters to form a raw time-series monitoring dataset;

[0007] Step 2: Based on the original time-series monitoring dataset, construct the time-domain drift feature channel and the component cross-interference feature channel to obtain a dual-channel state representation;

[0008] Step 3: Based on the preset historical steady-state baseline data, perform error source separation on the dual-channel state characterization to independently distinguish the current measurement deviation into zero-point drift component, sensitivity drift component and cross-interference component;

[0009] Step 4: Based on the zero-point drift component, sensitivity drift component, and cross-interference component, establish a dynamic coupling calibration model to eliminate drift bias and cross-sensitivity bias in real time, and output the compensated harmful gas concentration value.

[0010] Step 5: Based on the intermediate parameters of the calibration process, continuously evaluate the current compensation confidence level. When the compensation confidence level is lower than the preset threshold, trigger the standard gas verification prompt.

[0011] The present invention also provides a harmful gas concentration calibration system, the system comprising:

[0012] Data acquisition module: used to acquire multi-channel raw sensing signals of target harmful gases through gas sensing array, and simultaneously collect real-time environmental parameters to form raw time-series monitoring dataset;

[0013] Feature construction module: used to construct time-domain drift feature channel and component cross-interference feature channel based on the original time-series monitoring dataset, to obtain dual-channel state representation;

[0014] Error decoupling module: Based on preset historical steady-state baseline data, it is used to perform error source separation on the dual-channel state characterization, so as to independently distinguish the current measurement deviation into zero drift component, sensitivity drift component and cross interference component;

[0015] Dynamic calibration module: used to establish a dynamic coupling calibration model based on zero-point drift component, sensitivity drift component and cross-interference component, eliminate drift deviation and cross-sensitivity deviation in real time, and output the compensated harmful gas concentration value;

[0016] Confidence assessment module: Used to continuously assess the current compensation confidence level based on intermediate parameters of the calibration process. When the compensation confidence level is lower than a preset threshold, a standard gas verification prompt is triggered.

[0017] As can be seen from the above technical solution, the harmful gas concentration calibration method and system provided by the present invention have the following beneficial effects:

[0018] This invention addresses the shortcomings of existing technologies where three types of errors—sensor aging drift, multi-gas cross-sensitivity, and environmental temperature and pressure fluctuations—are treated together and cannot be independently identified and decoupled. By constructing a time-domain drift characteristic channel and a component cross-interference characteristic channel and performing error source separation based on historical steady-state baseline data, the current measurement deviation is directionally allocated along the physical source as a zero-point drift component, a sensitivity drift component, and a cross-interference component. This overcomes the technical contradiction of mutual aggravation of different error compensations under mixed processing. Under long-term operating conditions in industrial sites, it can significantly improve the accuracy retention capability of the compensated measurement values ​​and effectively reduce measurement errors in scenarios where multiple harmful gases coexist.

[0019] Furthermore, this invention continuously monitors the compensation amplitude of each compensation step through a confidence assessment step. When the compensation confidence is lower than a preset threshold, a standard gas verification prompt is triggered, which can promptly identify the risk of calibration model failure. At the same time, the standard gas verification is only triggered when the confidence is insufficient and only requires single-point verification. Compared with traditional solutions, this significantly reduces the frequency of standard gas consumption and the workload of on-site operation and maintenance, and achieves uninterrupted online calibration. Attached Figure Description

[0020] Other objects and results of the invention will become more apparent and readily understood by referring to the following description taken in conjunction with the accompanying drawings, and with a more complete understanding of the invention. In the drawings:

[0021] Figure 1 This is a schematic flowchart of a method for calibrating the concentration of harmful gases according to an embodiment of the present invention;

[0022] Figure 2 This is a functional block diagram of a harmful gas concentration calibration system according to an embodiment of the present invention. Detailed Implementation

[0023] In existing harmful gas concentration calibration schemes, measurement deviations from three different physical sources—sensor aging and drift, multi-gas cross-sensitivity, and environmental temperature and pressure fluctuations—are treated together and cannot be independently identified and decoupled. This leads to a rapid deterioration of calibration accuracy over time and a continuous accumulation of long-term measurement errors.

[0024] To address the aforementioned problems, this invention provides a method and system for calibrating hazardous gas concentrations. The core idea is as follows: By constructing a time-domain drift characteristic channel and a component cross-interference characteristic channel, the sensor output signal is mapped to an error characteristic space. Then, based on historical steady-state baseline data, the current measurement deviation is traced and separated to identify its source. The aliased overall deviation is directionally allocated along the physical source dimension into zero-point drift components, sensitivity drift components, and cross-interference components. Baseline zero-return correction, gain compensation, and interference subtraction are then performed on each component separately, and finally, the calibrated concentration value is output. Simultaneously, the compensation amplitude of each compensation step is continuously monitored during the calibration process to assess the compensation confidence level. When the confidence level is insufficient, a standard gas verification is triggered.

[0025] The essential difference between this invention and the prior art is that the prior art mixes errors from different physical sources together for unified correction, while this invention identifies and decouples errors independently according to their physical mechanisms, overcoming the technical contradiction of different error compensations aggravating each other. At the same time, it avoids the risk of data loss in unbounded adaptive calibration through a confidence assessment mechanism.

[0026] The following description of exemplary embodiments is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques and equipment should be considered part of the specification.

[0027] Reference Figure 1 The diagram shown is a schematic flowchart of a harmful gas concentration calibration method according to an embodiment of the present invention. In this embodiment, the harmful gas concentration calibration method includes:

[0028] Step 1: Acquire multi-channel raw sensing signals of the target harmful gas through a gas sensing array, and simultaneously collect real-time environmental parameters to form a raw time-series monitoring dataset.

[0029] This step is the data acquisition step and serves as the data entry point for the entire calibration method. A gas sensor array is deployed in the target monitoring area. The array contains multiple gas sensors of different types, each with varying degrees of sensitivity to different harmful gas components. The sensor array continuously outputs voltage or current response signals from each channel, which constitute the multi-channel raw sensor signals. Real-time environmental parameters, including temperature, humidity, and atmospheric pressure, are acquired synchronously through an environmental parameter acquisition module integrated near the sensor array.

[0030] Multi-channel raw sensor signals are continuously acquired at a set sampling frequency. The acquired signals are aligned with real-time environmental parameters using the same timestamp to form a raw time-series monitoring dataset. This dataset is stored in matrix form, with each row corresponding to a sampling time and each column corresponding to the response value of a sensor channel or an environmental parameter. This dataset serves as the data foundation for all subsequent processing steps.

[0031] Step 2: Based on the original time-series monitoring dataset, construct the time-domain drift feature channel and the component cross-interference feature channel to obtain a dual-channel state representation.

[0032] This step is the feature construction step. From the original time-series monitoring dataset, two independent feature channels are extracted along the time dimension and the component dimension, respectively.

[0033] The time-domain drift feature channel is used to characterize the slow changes in the response signal of each sensor channel over time. The specific construction method is as follows: First, the response signals of each sensor channel in the original time-series monitoring dataset are subjected to sliding window mean filtering to remove high-frequency random noise. The window width is preset according to the sensor response characteristics. The window slides point-by-point across the time-series data, and the response signals within the window are arithmetically averaged to obtain the smooth response curve of each sensor channel. Next, the smooth response curve is subtracted point-by-point from a preset clean air baseline value to obtain the time-domain drift trajectory sequence of each sensor channel. The clean air baseline value characterizes the standard response level of the sensor under conditions free of any target harmful gases or interfering gases. Finally, linear trend fitting is performed on the time-domain drift trajectory sequence to extract the monotonically changing drift trend term of each sensor channel over time, which serves as the time-domain drift feature channel. The slope of this drift trend term reflects the speed of sensor aging drift, and the intercept reflects the current cumulative drift amount.

[0034] The component cross-interference feature channel is used to characterize the cross-sensitive responses of coexisting components in the gas mixture to each sensing channel. The specific construction method is as follows: Based on the response amplitudes of each sensing channel at the same time in the original time-series monitoring dataset, the ratio of response amplitudes between any two sensing channels is calculated, generating a ratio vector sequence. This ratio vector sequence reflects the selectivity differences of each sensor to different gas components. A pre-constructed standard response ratio template library for each single harmful gas component is retrieved. This template library was obtained by testing and calibrating each single component gas under laboratory conditions beforehand. Each template in the library corresponds to the response ratio vector of one gas component on each sensing channel. The real-time calculated ratio vector sequence is matched with each template in the template library for similarity. The gas component corresponding to the template with the highest similarity is used to determine the interfering component present in the current gas mixture. The relative concentration percentage of the interfering component is determined based on the quantified value of the matching metric. The interfering component identifier and relative concentration percentage are used as the component cross-interference feature channel.

[0035] The time-domain drift characteristic channel and the component cross-interference characteristic channel together constitute a dual-channel state characterization. The former tracks the long-term degradation state of the sensor itself, while the latter tracks the component composition state of the gas environment to be measured. The establishment of the dual-channel state characterization enables subsequent steps to distinguish the sources of error at the physical mechanism level.

[0036] Step 3: Based on the preset historical steady-state baseline data, perform error source separation on the dual-channel state characterization to independently distinguish the current measurement deviation into zero-point drift component, sensitivity drift component and cross-interference component.

[0037] This step is the error decoupling step. Historical steady-state baseline data consists of standard response records of the gas sensor array during its historical operation under clean air conditions and known concentrations of calibration gas. The historical steady-state baseline data includes three parts: the standard zero-point baseline, i.e., the standard response value of each sensor channel under conditions without any target harmful gas; the initial response amplitude, i.e., the standard response amplitude of each sensor channel to a known concentration of calibration gas during factory calibration; and the unit concentration response coefficient, i.e., the sensitivity coefficient of each sensor channel to each individual gas component.

[0038] The current measurement bias is the combined difference vector between the dual-channel state characterization and the historical steady-state baseline data. This step directs the distribution of this combined difference along three physical sources.

[0039] The zero-point drift component is separated as follows: First, multiple historical response data segments of the gas sensor array during the period of no target harmful gas introduction are extracted, and the mean of each historical response data segment is calculated to obtain multiple sets of zero-point baseline reference values. The period of no target harmful gas introduction is pre-calibrated according to the on-site operating conditions, such as during process breaks or system self-testing periods. Then, the multiple sets of zero-point baseline reference values ​​are arranged in chronological order of acquisition and smoothed using an exponentially weighted moving average method to reduce the impact of random fluctuations, obtaining the estimated zero-point baseline value at the current moment. The smoothing coefficient of the exponentially weighted moving average method is preset according to the sensor drift characteristics. Finally, the difference between the estimated zero-point baseline value and the corresponding standard zero-point baseline in the historical steady-state baseline data is taken as the zero-point drift component. The zero-point drift component characterizes the slow shift of the sensor baseline over time.

[0040] The sensitivity drift component is separated as follows: First, the historical response amplitude sequence of the gas sensor array to a known concentration of calibration gas is obtained, along with the corresponding initial response amplitude from the historical steady-state baseline data. The historical response amplitude sequence records multiple responses of the sensor during the calibration gas introduction. Then, the ratio of the response amplitude at each moment in the historical response amplitude sequence to the initial response amplitude is calculated to obtain the sensitivity decay rate sequence. A ratio less than 1 indicates sensitivity decay, while a ratio greater than 1 indicates sensitivity enhancement. Finally, linear regression fitting is performed on the sensitivity decay rate sequence to obtain an estimated sensitivity decay rate at the current moment, which serves as the sensitivity drift component. The sensitivity drift component characterizes the degree of change in the sensor's sensitivity to the target gas.

[0041] The separation of cross-interference components is as follows: First, the identifier of the currently matched interfering component and its corresponding relative concentration percentage are extracted from the component cross-interference feature channel. This interfering component identifier represents other gas components coexisting in the current gas mixture besides the target harmful gas. Then, based on the interfering component identifier, the unit concentration response coefficient of the corresponding interfering component in each sensor channel is retrieved from the historical steady-state baseline data. The unit concentration response coefficient characterizes the response amplitude generated by the interfering component at a unit concentration in each sensor channel. Finally, the relative concentration percentage is multiplied by the unit concentration response coefficient of each sensor channel to obtain the additional response amplitude contributed by the corresponding interfering component in each sensor channel, which is taken as the cross-interference component. The cross-interference component characterizes the degree of interference of the coexisting gas on the measurement of the target harmful gas concentration.

[0042] Through the above processing, the zero-point drift component, sensitivity drift component, and cross-interference component correspond to three different physical sources: long-term aging of the sensor, sensitivity degradation, and gas cross-interference, respectively, thus achieving independent characterization of the error.

[0043] Step 4: Based on the zero-point drift component, sensitivity drift component, and cross-interference component, establish a dynamic coupling calibration model to eliminate drift bias and cross-sensitivity bias in real time, and output the compensated harmful gas concentration value.

[0044] This step is a dynamic calibration step. The dynamic coupling calibration model consists of three compensation steps executed sequentially.

[0045] The first compensation stage performs baseline zero-return correction on the original multi-channel sensor signals based on the zero-point drift component. Specifically, the zero-point drift component is subtracted from the original response signal of each sensor channel to obtain the zero-point compensated sensor signal. This stage eliminates the measurement error caused by sensor baseline offset.

[0046] The second compensation stage performs gain compensation on the zero-point compensated sensing signal based on the sensitivity drift component. The sensitivity drift component is characterized by an estimated sensitivity attenuation rate. Gain compensation is achieved by multiplying the zero-point compensated sensing signal by the reciprocal of the estimated sensitivity attenuation rate to obtain the sensitivity-compensated sensing signal. This stage restores the sensitivity lost by the sensor due to aging.

[0047] The third compensation stage performs interference subtraction on the sensitivity-compensated sensor signal based on the cross-interference component. Specifically, it subtracts the additional response amplitude corresponding to the cross-interference component from the sensitivity-compensated sensor signal to obtain the interference-compensated sensor signal. This stage eliminates the cross-interference of coexisting gases on the target gas measurement.

[0048] The interference-compensated sensor signal is substituted into a preset concentration inversion function to obtain the calibrated harmful gas concentration value. The concentration inversion function is a function pre-established based on the quantitative relationship between the sensor response value and the gas concentration. The concentration inversion function adopts a cubic polynomial fitting form, and its expression is:

[0049] ;

[0050] in, This is a numerical value for the calibrated harmful gas concentration. It only represents the quantitative relationship between the sensor response and the gas concentration; the value is a dimensionless number. The coefficient is the dimensionless value of the sensor signal after interference compensation. , , , denoted as the fitting coefficients determined by least squares fitting, where Characterizing the nonlinear curvature of the sensor response curve, Characterizing the degree of nonlinear bending of the response curve, Characterize the linear gain of the sensor in the small signal region. The intercept of the fitted curve.

[0051] The above fitting coefficients were pre-calibrated in the following way: Under laboratory conditions, gas was sequentially introduced into the gas sensing array. A set of target gas standard samples with known concentrations, the concentration values ​​of which are denoted as follows: Record the sensor channel response values ​​corresponding to each concentration point. ,form Group concentration-response data pair ,in ,and A value of at least 4 is required to ensure sufficient degrees of freedom in the fit. Based on this... Data pairs, in terms of concentration values Dependent variable, response value Using the least squares method as the independent variable, a cubic polynomial fitting function is constructed to achieve the sum of squares of the fitting errors. To reach the minimum, the coefficient is determined accordingly. , , , The value.

[0052] In the actual calibration process, the dimensionless value of the interference-compensated sensor signal is used as the independent variable. Substituting the values ​​into the cubic polynomial above, the corresponding concentration values ​​can be calculated. This outputs the calibrated concentration of harmful gases. Once calibrated, this concentration inversion function is stored in the system for concentration calculations during subsequent real-time calibration processes.

[0053] Step 5: Continuously evaluate the current compensation confidence level based on the intermediate parameters of the calibration process. When the compensation confidence level is lower than the preset threshold, trigger the standard gas verification prompt.

[0054] This step is a confidence assessment step, used to monitor the reliability of the calibration model.

[0055] The assessment method for compensation confidence is as follows: Obtain the compensation magnitude of each compensation step in the dynamic coupling calibration model. The compensation magnitude includes zero-point correction, gain adjustment, and interference deduction. The zero-point correction is the actual baseline offset correction value applied in the first compensation stage, the gain adjustment is the actual amplification factor adjustment value applied in the second compensation stage, and the interference deduction is the actual additional response magnitude deducted in the third compensation stage.

[0056] Each compensation value is compared with the corresponding steady-state fluctuation range in the historical steady-state baseline data. The steady-state fluctuation range of the compensation value is predetermined based on the statistical distribution of the compensation value during the sensor's historical normal operation. The number of compensation items exceeding their respective steady-state fluctuation ranges is counted, specifically the number of zero-point correction, gain adjustment, and interference deduction items exceeding their corresponding steady-state fluctuation ranges. The current compensation confidence level is determined based on the proportion of compensation items exceeding the steady-state fluctuation range to the total number of compensation items. A higher proportion indicates a greater deviation of the calibration model from the normal state, and a lower current compensation confidence level.

[0057] When the compensation confidence level falls below a preset threshold, a standard gas verification prompt is triggered. The specific process for standard gas verification is as follows: A standard gas verification command is generated and sent to the control interface of the standard gas delivery device. After receiving the command, the standard gas delivery device introduces a standard gas at a single concentration point into the gas sensor array. The response value of the gas sensor array to the standard gas is obtained, and this response value is compared with the nominal concentration value of the standard gas to calculate the verification deviation rate. The verification deviation rate reflects the degree of deviation between the current calibration result and the standard gas reference. Based on the verification deviation rate, the compensation coefficients of the dynamic coupling calibration model are adjusted, for example, by adjusting the compensation coefficients of the gain compensation stage or the parameters of the concentration inversion function, thereby correcting the systematic bias of the calibration model.

[0058] It should be noted that the compensation process described in steps 1-4 of this invention does not require the intervention of standard gas under normal monitoring conditions; it dynamically corrects the original sensor signal based on the algorithm model. The "standard gas verification" in step 5 is triggered only when the model confidence level is insufficient. Its purpose is to obtain on-site reference samples to correct the algorithm model coefficients, ensuring that the algorithm compensation result maintains traceability consistency with the legally mandated calibration value. This verification is not equivalent to the periodic mandatory calibration stipulated in national metrological regulations, but rather an auxiliary maintenance method of this compensation method.

[0059] In one feasible implementation, this embodiment uses an online monitoring system for hazardous gases at the boundary of a petrochemical plant as an example. The system is deployed at the downwind boundary of the plant area to monitor the concentrations of hydrogen sulfide and sulfur dioxide in the air in real time. The gas sensor array includes an electrochemical hydrogen sulfide sensor, an electrochemical sulfur dioxide sensor, and temperature, humidity, and air pressure sensors. Each sensor continuously collects data at a sampling frequency of once per second. Historical steady-state baseline data was obtained through clean air environment testing and standard gas calibration before the system was put into operation and is stored in the local data storage unit.

[0060] The system operated normally during the initial commissioning phase. After 90 days of operation, the electrolyte in the electrochemical sensor was gradually consumed, and the surface of the sensitive electrode became passivated. Under these conditions, the acquisition step obtained raw sensor signals from multiple channels and real-time environmental parameters to form a raw time-series monitoring dataset. The feature construction step extracted the time-domain drift trajectory sequences of each sensor channel from the dataset. Linear trend fitting identified the drift trend term slope for the hydrogen sulfide channel as 0.15 mV per day and the drift trend term slope for the sulfur dioxide channel as 0.08 mV per day. The time-domain drift feature channel outputs the drift trend term. Simultaneously, the ratio vector sequence was matched with a template library to identify no significant interfering components in the current environment. The component cross-interference feature channel outputs an indicator of no interfering components.

[0061] The error decoupling step is based on historical steady-state baseline data for error source tracing and separation. In the separation of the zero-point drift component, four segments of historical response data from the system during the early morning process shutdown are extracted. The mean values ​​of each segment are 2.08 mV, 2.12 mV, 2.06 mV, and 2.10 mV, respectively. After smoothing with an exponentially weighted moving average, the estimated zero-point baseline value at the current time is 2.09 mV, which differs from the standard zero-point baseline of 2.00 mV by 0.09 mV. The zero-point drift component is +0.09 mV. In the separation of the sensitivity drift component, the historical response amplitudes of the system to the known concentration of calibration gas on days 30, 60, and 90 were obtained as 42.5 mV, 41.2 mV, and 40.1 mV, respectively, and the initial response amplitude was 45.0 mV. The ratio of the response amplitude to the initial response amplitude at each time point was calculated to obtain the sensitivity decay rate sequence as 0.944, 0.916, and 0.891. The sensitivity decay rate estimate extrapolated to the current time point by linear regression fitting is 0.887, and the sensitivity drift component is 0.887.

[0062] The dynamic calibration process involves three compensation steps. The first step performs baseline zeroing correction on the original hydrogen sulfide channel response signal based on a zero-point drift component of +0.09 mV. The second step performs gain compensation on the zero-point compensated signal based on a sensitivity drift component of 0.887, with an amplification factor of 1 divided by 0.887 equaling 1.127. The third step confirms the absence of significant interfering components through cross-interference characteristic channel matching, resulting in zero interference subtraction. The interference-compensated signal is then substituted into the concentration inversion function, a cubic polynomial based on least-squares fitting, outputting a calibrated hydrogen sulfide concentration of 0.98 mg / m³.

[0063] The confidence assessment step yielded a zero-point correction of 0.09 mV, a gain adjustment of 1.127, and an interference deduction of 0. The steady-state fluctuation ranges for each compensation value were 0.00 to 0.05 mV for the zero-point correction, 1.00 to 1.05 mV for the gain adjustment, and -0.02 to 0.02 mV for the interference deduction. Both the zero-point correction and gain adjustment exceeded their steady-state fluctuation ranges. The number of compensation items exceeding these ranges was 2, and the total number of compensation items was 3, representing 66.7%. A preset threshold of 50% was set; the current compensation confidence level was lower than the preset threshold, triggering a standard gas review prompt.

[0064] During the standard gas verification process, the standard gas delivery device introduced hydrogen sulfide standard gas with a concentration of 1.00 mg / m³ into the sensor array. After processing by the current calibration model, the sensor response value was displayed as 0.87 mg / m³, with a verification deviation rate of 13%. Based on the verification deviation rate, the compensation coefficient of the dynamic coupling calibration model was adjusted, and the amplification factor of the gain compensation stage was corrected from 1.127 to 1.127 multiplied by 1.13, which equals 1.274, thus completing the calibration model correction.

[0065] Before adopting this method, the deviation between the measured value and the standard gas on the 90th day under the same operating conditions was 15.2%. After adopting this method, the deviation of the calibration output was reduced to 1.3%, effectively restoring the measurement accuracy. In the subsequent 30 days of continuous operation, the deviation between the calibration output concentration value and the standard gas comparison value obtained from periodic sampling remained within ±3%.

[0066] In one feasible implementation, this embodiment uses a multi-gas monitoring scenario in a coal mine as an example. The monitoring point is located in the return airway of the mining face and needs to simultaneously monitor methane and carbon monoxide concentrations. The gas sensor array includes a catalytic combustion methane sensor, an electrochemical carbon monoxide sensor, and temperature, humidity, and pressure sensors. In addition to the target gas methane, the underground environment also contains small amounts of carbon monoxide and carbon dioxide, with carbon monoxide causing positive cross-interference with the catalytic combustion methane sensor. Historical steady-state baseline data was obtained during system installation.

[0067] During system operation, the acquisition step obtains multi-channel raw sensor signals and real-time environmental parameters to form a raw time-series monitoring dataset. In the feature construction step, the ratio vector sequence is matched with a template library to identify carbon monoxide as an interfering component in the current gas mixture, in addition to methane. The matching similarity is 0.92, and the relative concentration percentage is 0.3%. The component cross-interference feature channel outputs the interfering component as carbon monoxide with a relative concentration percentage of 0.3%. The time-domain drift feature channel outputs the drift trend terms for each channel.

[0068] In the error decoupling step, the zero-point drift component separation yielded a zero-point drift component of 0.02 mV for the methane channel and 0.01 mV for the carbon monoxide channel. Sensitivity drift component separation revealed a sensitivity decay rate of 0.950 for the methane channel and 0.960 for the carbon monoxide channel. In the cross-interference component separation, the interfering component, identified as carbon monoxide and its relative concentration of 0.3%, was extracted from the cross-interference characteristic channel. The unit concentration response coefficient of carbon monoxide in the methane sensing channel was retrieved from historical steady-state baseline data, yielding 0.25 mV per percentage. The additional response amplitude of carbon monoxide to the methane channel was calculated to be 0.3% multiplied by 0.25 mV per percentage, equaling 0.075 mV, which was taken as the cross-interference component.

[0069] In the dynamic calibration process, the first step performs zero-point homing correction on the methane channel, subtracting the zero-point drift component of 0.02 mV. The second step performs gain compensation on the zero-point compensated signal, multiplying it by an amplification factor of 1 and dividing by 0.950, which equals 1.053. The third step subtracts the cross-interference component of 0.075 mV from the sensitivity-compensated signal. Substituting the interference-compensated signal into the concentration inversion function, the output calibrated methane concentration is 0.52% volume fraction. Without this calibration, the methane concentration calculated solely from the sensor's original response value is approximately 0.68% volume fraction, a deviation of 33.3% compared to the actual value of 0.51% volume fraction, posing a risk of false alarms. After calibration using this method, the deviation is reduced to 2.0%.

[0070] The confidence assessment step obtained the amplitude of each compensation amount, all of which were within the steady-state fluctuation range, with zero items exceeding the range. The compensation confidence level was 100%, and the standard gas review was not triggered. The system runs continuously online, with uninterrupted monitoring.

[0071] like Figure 2 The diagram shown is a functional block diagram of a hazardous gas concentration calibration system 100 provided in an embodiment of the present invention, including a data acquisition module 101, a feature construction module 102, an error decoupling module 103, a dynamic calibration module 104, and a confidence assessment module 105.

[0072] In this embodiment, the functions of each module are as follows:

[0073] The data acquisition module 101 is used to acquire multi-channel raw sensor signals of the target harmful gas through a gas sensor array, and simultaneously acquire real-time environmental parameters to form a raw time-series monitoring dataset. The hardware deployment of the data acquisition module includes a gas sensor array and an environmental parameter acquisition module, which are integrated into the field monitoring terminal. The analog signals are converted into digital signals through an analog-to-digital converter interface and then transmitted to the data processing unit.

[0074] The feature construction module 102 is used to construct a time-domain drift feature channel and a component cross-interference feature channel based on the original time-series monitoring dataset, thereby obtaining a dual-channel state representation. This module internally includes a time-domain drift feature extraction submodule and a component cross-interference feature extraction submodule, which respectively perform operations such as sliding window mean filtering, point-by-point difference calculation, linear trend fitting, ratio vector calculation, and template library matching.

[0075] The error decoupling module 103 is used to perform error source separation on the dual-channel state characterization based on preset historical steady-state baseline data, and to independently distinguish the current measurement deviation into zero-point drift component, sensitivity drift component, and cross-interference component. This module contains a zero-point drift separation submodule, a sensitivity drift separation submodule, and a cross-interference separation submodule, and each submodule executes its own component extraction logic in parallel or sequentially.

[0076] The dynamic calibration module 104 is used to establish a dynamic coupling calibration model based on the zero-point drift component, sensitivity drift component, and cross-interference component, to eliminate drift bias and cross-sensitivity bias in real time, and to output the compensated harmful gas concentration value. This module internally includes a baseline zero-return correction submodule, a gain compensation submodule, and an interference subtraction submodule, which are executed in cascade.

[0077] The confidence assessment module 105 is used to continuously assess the current compensation confidence level based on intermediate parameters of the calibration process, and triggers a standard gas verification prompt when the compensation confidence level is lower than a preset threshold. This module internally includes a compensation amount acquisition submodule, a steady-state interval comparison submodule, a confidence level determination submodule, and a standard gas verification control submodule.

[0078] The above modules are deployed in the intelligent converged terminal of the distribution area or the local edge computing gateway. They are connected to the gas sensor array, environmental parameter acquisition module and standard gas delivery device through the bus interface to realize real-time data acquisition, online calibration and confidence monitoring.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for calibrating the concentration of harmful gases, characterized in that, The method includes: Step 1: Acquire multi-channel raw sensing signals of the target harmful gas through a gas sensing array, and simultaneously collect real-time environmental parameters to form a raw time-series monitoring dataset; Step 2: Based on the original time-series monitoring dataset, construct the time-domain drift feature channel and the component cross-interference feature channel to obtain a dual-channel state representation; Step 3: Based on the preset historical steady-state baseline data, perform error source separation on the dual-channel state characterization to independently distinguish the current measurement deviation into zero-point drift component, sensitivity drift component and cross-interference component; Step 4: Based on the zero-point drift component, sensitivity drift component, and cross-interference component, establish a dynamic coupling calibration model to eliminate drift bias and cross-sensitivity bias in real time, and output the compensated harmful gas concentration value. Step 5: Based on the intermediate parameters of the calibration process, continuously evaluate the current compensation confidence level. When the compensation confidence level is lower than the preset threshold, trigger the standard gas verification prompt.

2. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The construction of the temporal drift feature channel includes: The response signals of each sensor channel in the original time-series monitoring dataset are subjected to sliding window mean filtering to remove high-frequency random noise and obtain smooth response curves for each sensor channel. The time-domain drift trajectory sequence of each sensor channel is obtained by performing point-by-point difference calculation between the smooth response curve and the preset clean air baseline value. Linear trend fitting is performed on the time-domain drift trajectory sequence to extract the drift trend term that changes monotonically with time for each sensing channel, which is used as the time-domain drift feature channel.

3. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The constructed component cross-interference feature channel includes: Based on the response amplitude of each sensor channel in the original time-series monitoring dataset, calculate the ratio of response amplitude between any two sensor channels and generate a ratio vector sequence. Retrieve a pre-constructed standard response ratio template library for each individual hazardous gas component; The ratio vector sequence is matched with each template in the standard response ratio template library for similarity. Based on the matching results, the interfering components and their corresponding relative concentrations in the current gas mixture are determined and used as the component cross-interference feature channel.

4. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The step of independently classifying the current measurement deviation into zero-point drift components includes: Multiple historical response data segments of the gas sensing array during the period of introduction of no target harmful gas were extracted, and the mean value of each historical response data segment was calculated to obtain multiple sets of zero-point baseline reference values. Multiple sets of zero-point baseline reference values ​​are arranged in chronological order of acquisition time, and smoothed using the exponential weighted moving average method to obtain the zero-point baseline estimate at the current time. The difference between the zero-point baseline estimate and the corresponding standard zero-point baseline in the preset historical steady-state baseline data is taken as the zero-point drift component.

5. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The step of independently classifying the current measurement deviation into sensitivity drift components includes: Obtain the historical response amplitude sequence of the gas sensing array to a known concentration of calibration gas, as well as the corresponding initial response amplitude in the historical steady-state baseline data; Calculate the ratio of the response amplitude at each time point in the historical response amplitude sequence to the initial response amplitude to obtain the sensitivity decay rate sequence; Linear regression fitting is performed on the sensitivity decay rate sequence to obtain the estimated sensitivity decay rate at the current moment, which is used as the sensitivity drift component.

6. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The step of independently classifying the current measurement deviation into cross-interference components includes: Extract the identifiers of the currently successfully matched interfering components and their corresponding relative concentration percentages from the component cross-interference feature channels; Based on the identifier of the interfering component, retrieve the unit concentration response coefficient of the corresponding interfering component in each sensing channel from the historical steady-state baseline data. Multiplying the relative concentration percentage by the unit concentration response coefficient of each sensing channel yields the additional response amplitude contributed by the corresponding interference component on each sensing channel, which is taken as the cross-interference component.

7. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The establishment of the dynamic coupling calibration model includes: Based on the zero-point drift component, baseline zero-return correction is performed on the original multi-channel sensing signals to obtain the zero-point compensated sensing signals. Based on the sensitivity drift component, gain compensation is performed on the zero-point compensated sensing signal to obtain the sensitivity compensated sensing signal. Based on the cross-interference component, interference subtraction is performed on the sensitivity-compensated sensing signal. The additional response amplitude corresponding to the cross-interference component is subtracted to obtain the interference-compensated sensing signal. The interference-compensated sensor signal is substituted into the preset concentration inversion function to obtain the calibrated harmful gas concentration value.

8. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The continuous assessment of the current compensation confidence level includes: Obtain the compensation magnitude of each compensation step in the dynamic coupling calibration model. The compensation magnitude includes zero-point correction, gain adjustment, and interference deduction. Compare the magnitude of each compensation amount with the corresponding steady-state fluctuation range of the compensation amount in the historical steady-state baseline data, and count the number of compensation items that exceed their respective steady-state fluctuation ranges. The current compensation confidence level is determined by the proportion of the number of compensation items that exceed the steady-state fluctuation range to the total number of compensation items. The higher the proportion, the lower the current compensation confidence level.

9. The method for calibrating the concentration of harmful gases as described in claim 1, characterized in that, The triggering prompt for standard gas verification includes: When the current compensation confidence level is lower than the preset threshold, a standard gas verification instruction is generated; According to the standard gas verification command, control the standard gas delivery device to introduce standard gas at a single concentration point into the gas sensor array; The response value of the gas sensor array to the standard gas is obtained, and the response value is compared with the nominal concentration value of the standard gas to obtain the verification deviation rate. Adjust the compensation coefficients of the dynamic coupling calibration model based on the verification deviation rate.

10. A harmful gas concentration calibration system, characterized in that, The system for implementing the harmful gas concentration calibration method according to any one of claims 1-9, the system comprising: Data acquisition module: used to acquire multi-channel raw sensing signals of target harmful gases through gas sensing array, and simultaneously collect real-time environmental parameters to form raw time-series monitoring dataset; Feature construction module: used to construct time-domain drift feature channel and component cross-interference feature channel based on the original time-series monitoring dataset, to obtain dual-channel state representation; Error decoupling module: Based on preset historical steady-state baseline data, it is used to perform error source separation on the dual-channel state characterization, so as to independently distinguish the current measurement deviation into zero drift component, sensitivity drift component and cross interference component; Dynamic calibration module: used to establish a dynamic coupling calibration model based on zero-point drift component, sensitivity drift component and cross-interference component, eliminate drift deviation and cross-sensitivity deviation in real time, and output the compensated harmful gas concentration value; Confidence assessment module: Used to continuously assess the current compensation confidence level based on intermediate parameters of the calibration process. When the compensation confidence level is lower than a preset threshold, a standard gas verification prompt is triggered.

Images