In-situ zero-point self-calibration method for electrochemical ammonia gas sensor incorporating environmental parameters

Abstract

This invention belongs to the field of self-calibration technology for electrochemical gas sensors, and discloses an in-situ zero-point self-calibration method for an electrochemical ammonia sensor that integrates environmental parameters. The method includes: a controller calculating the first derivatives of environmental temperature and humidity parameters; identifying a baseline transition risk event when the absolute value of the first derivative exceeds a preset environmental disturbance threshold; responding to the event, the controller maintains its adjustment output signal for the controlled object at the output value before the event identification and initiates a stabilization window; based on the maintained adjustment output signal, the controller establishes a quasi-static technical premise for the process, and constrained by this premise, determines the difference in sensor measurement values ​​before and after the stabilization window as the baseline step value, which is used to update the cumulative zero-point offset state variable. This invention enables the control system to autonomously identify baseline step pseudo-signals generated by the sensor due to environmental transients and actively intervenes to prevent interference with the control logic.

Description

Technical Field

[0001] This invention relates to an in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor that integrates environmental parameters, belonging to the field of self-calibration technology for electrochemical gas sensors. Background Technology

[0002] Currently, electrochemical ammonia sensors are widely used in environmental monitoring, livestock and poultry farming, industrial safety, and process control due to their high sensitivity, fast response, and relatively low cost. However, their core working principle determines that the sensor's output baseline will inevitably drift slowly over time, making it susceptible to interference from fluctuations in ambient temperature and humidity. To address this issue, the mainstream solution in the industry is to periodically perform manual calibration or attempt to establish a static compensation model based on the absolute values ​​of ambient temperature and humidity. This static compensation design approach generally regards zero-point drift as a smooth and continuous analog quantity offset that changes with the absolute values ​​of ambient temperature and humidity. Therefore, the technical solutions all focus on establishing a static functional relationship, attempting to calculate the corresponding drift amount in real time by measuring temperature and humidity values ​​and then subtracting it.

[0003] This approach overlooks the key characteristics of the drift behavior of electrochemical sensors when installed in situ in real industrial settings: discontinuity and dynamic hysteresis. In real industrial processes or complex environmental monitoring, environmental parameters, especially humidity and temperature, often change drastically and transiently, such as process start-up and shutdown, instantaneous evaporation of condensate, or environmental airflow disturbances. When responding to such transient environmental shocks, the physicochemical state of the electrolyte or sensitive electrode surface of the electrochemical sensor will undergo a brief period of disequilibrium. After re-reaching equilibrium, the measurement baseline is likely to experience a step change. The amplitude and direction of this baseline step change are highly correlated with the rate of change of environmental parameters, not just their absolute values. Traditional static compensation models are, in principle, unable to predict and handle such steady-state transitions triggered by dynamic processes.

[0004] For example, the ammonia injection controller in an SCR denitrification system cannot distinguish between a genuine sudden change in ammonia concentration in the process and a baseline jump in the sensor itself. The system may misinterpret this as a serious process disturbance and make incorrect adjustments, such as drastically increasing or decreasing the ammonia injection rate. This can lead to severe overshoot or instability in the control loop, seriously affecting the stability, economy, and safety of the main process. To address these issues, in addition to the mainstream static compensation model, there are data-driven modeling approaches that attempt to identify sensor offsets through software algorithms. For instance, Chinese invention patent CN119164436A discloses a self-calibration method and related device for a micro air quality monitoring station. However, the calibration steps rely on specific calibration conditions, such as a pollution-free environment or an environment with known concentrations. This is not feasible in real-world, continuously operating industrial process control, such as SCR denitrification, and it fails to solve the technical challenge of real-time closed-loop control interference from sensor baseline step pseudo-signals.

[0005] Therefore, the technical problem to be solved by this invention is how to enable the control system to autonomously identify the sensor baseline step signal caused by environmental transients, avoid interference with its own control logic, and seize the opportunity to complete adaptive tracking of the new baseline. Summary of the Invention

[0006] This invention provides an in-situ zero-point self-calibration method for an electrochemical ammonia sensor that integrates environmental parameters. Its main purpose is to solve the problem of how the control system can autonomously identify the sensor baseline step signal caused by environmental transients, avoid interference from its own control logic, and take the opportunity to complete the adaptive tracking of the new baseline.

[0007] To achieve the above objectives, this invention provides an in-situ zero-point self-calibration method for an electrochemical ammonia sensor that integrates environmental parameters. This method is applied to a control system including an electrochemical ammonia sensor, an environmental parameter acquisition unit, and a controller. The method includes:

[0008] The controller acquires the sensor measurements from the electrochemical ammonia sensor, as well as the ambient temperature and humidity parameters from the environmental parameter acquisition unit; the controller calculates the first derivatives of the ambient temperature and humidity parameters; the controller compares the absolute value of the first derivative with a preset environmental disturbance threshold to identify baseline transition risk events.

[0009] In response to the identification of a baseline transition risk event, the controller executes the following steps: Step a, maintaining its adjustment output signal for the controlled object at the output value just before the event identification, and storing the first sensor measurement value just before the event identification; Step b, starting a stabilization window, monitoring the sensor measurement values ​​within this window until the rate of change of the sensor measurement values ​​is lower than a preset signal stabilization threshold, then acquiring the second sensor measurement value at this moment, and the controller then executes the following steps: Step c, based on the fact that the adjustment output signal has been maintained at the output value in step a, establishing a quasi-static technical premise for the process of the controlled object from the moment before the event identification to the moment of acquiring the second sensor measurement value; Step d, constrained by the quasi-static technical premise, determining the difference between the second sensor measurement value and the first sensor measurement value as the step amount of the sensor zero-point baseline caused by environmental transients; Step e, accumulating the step amount into the cumulative zero-point offset state variable;

[0010] After the stabilization window ends, the controller releases the maintenance of the regulation output signal and resumes closed-loop regulation of the controlled object based on the difference between the sensor measurement value and the updated cumulative zero-point offset state variable.

[0011] Preferably, in step b, the stabilization window has a preset maximum duration; if, within the stabilization window, the rate of change of the sensor measurement value is lower than the preset signal stability threshold and remains stable for a preset time before reaching the maximum duration, the controller ends the stabilization window early and acquires the second sensor measurement value; if, until the end of the maximum duration, the rate of change is still not lower than the preset signal stability threshold, the controller reads the sensor reading at the end of the maximum duration as the second sensor measurement value.

[0012] Preferably, the first derivative is obtained in real time by the controller through high-frequency oversampling of the ambient temperature and humidity parameters, and by continuous difference operation or digital filtering algorithm.

[0013] Preferably, the method further includes: the controller establishing and maintaining an environmental modulation decoupling model, which describes the intrinsic modulation relationship between environmental temperature and humidity parameters on the sensor zero-point baseline; after step d, the controller further performs: using the environmental modulation decoupling model, predicting the model step based on the environmental parameters at the moment before event identification and the moment when the second sensor measurement value is obtained; calculating the verification deviation between the difference and the model step, and recording the verification deviation in association with the calibration event of step e.

[0014] Preferably, the method further includes: the controller also performs a clean window identification step, which includes: monitoring the time series pattern of sensor measurements and environmental parameters; identifying a clean window when the sensor measurements are within a preset historical data statistical low percentile interval and their fluctuation variance is lower than a preset stability threshold and the environmental parameters are within a preset typical range; during the clean window period, the controller acquires the sensor measurements and the corresponding environmental temperature and humidity parameters, and uses the sensor measurements and the corresponding environmental temperature and humidity parameters to correct the cumulative zero-point offset state variable or to calibrate the environmental modulation decoupling model online.

[0015] Preferably, the clean window identification step further includes: the controller introducing a confidence assessment mechanism; the controller comprehensively assesses the duration of the clean window, the data stability of the sensor measurements, and the degree of agreement with the prediction of the environmental modulation decoupling model, and generates a confidence score; only when the confidence score exceeds a preset confidence threshold will the controller perform correction of the cumulative zero-point offset state variable or online calibration of the environmental modulation decoupling model.

[0016] Preferably, when the controller performs step e or performs correction, it adopts a progressive update strategy. The progressive update strategy includes: after the controller calculates the step amount or correction amount, it does not apply it to the cumulative zero-point offset state variable all at once, but smoothly updates the step amount or correction amount to the cumulative zero-point offset state variable over multiple control cycles using a preset weighted average algorithm or by increasing it in preset small steps.

[0017] Preferably, the method further includes a dual verification and backtracking arbitration mechanism for the calibration action. The mechanism includes: after executing step e, the controller immediately enters the post-hoc monitoring period to check whether the short-term stability and response characteristics of the sensor measurements near the new baseline are within a preset reasonable range; the controller records the calibration event and its context data of step e as a traceable log; if the calibration event is found to cause abnormal data behavior during the post-hoc monitoring period or in subsequent operation, the controller initiates arbitration and automatically rolls back the accumulated zero-point offset state variable to the state before the calibration event.

[0018] Preferably, the controlled object is an ammonia injection actuator in a selective catalytic reduction (SCR) denitrification system, and the adjusted output signal is the control signal of the ammonia injection valve. The controller maintains the control signal of the ammonia injection valve at the output value just before the event recognition through step a, so as to avoid incorrect adjustment of the control signal of the ammonia injection valve due to the step amount of the sensor zero-point baseline during the stabilization window, thereby maintaining the stable operation of the SCR denitrification system and preventing ammonia escape.

[0019] Preferably, the controller also deploys an adaptive learning framework; the adaptive learning framework is used to continuously absorb operational data, including the identification record of baseline transition risk events, the calibration record of step e, the identification record of clean window, the verification deviation, and the execution record of traceable logs and arbitration; the controller periodically uses the operational data to iteratively optimize the environmental disturbance threshold, the signal stability threshold, the preset historical data statistical low percentile interval, the preset typical range, or the internal parameters of the environmental modulation decoupling model.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. By identifying the transient change characteristics of environmental parameters, the system can predict possible step changes in the sensor measurement baseline. This mechanism enables the control system to identify the sudden change signal generated by the sensor as a non-process disturbance when encountering such environmental shocks, thus preventing the control system from executing incorrect adjustment actions due to misjudged signals and ensuring the smooth operation of the main control process.

[0022] 2. After identifying a transient change in the environment, the controller actively freezes its adjustment output and simultaneously starts a stabilization window. The freezing behavior creates quasi-static conditions for the process in the control logic, while the stabilization process waits for the sensor to reach electrochemical equilibrium in the new environment. The combination of these two factors allows the controller to logically attribute the signal difference observed during this window period to the baseline step of the sensor itself, thus achieving in-situ self-calibration.

[0023] 3. This invention provides an event-driven baseline tracking method that does not rely on establishing a static functional relationship between complex environmental parameters and drift. Instead, it captures and accumulates the relative baseline difference caused by transients through active intervention of the controller when environmental disturbances actually occur. This method frees zero-point calibration from dependence on physical standard gases. The controller achieves continuous dynamic tracking of the sensor's true baseline by updating the internally maintained zero-point offset state variable. It transforms the sensor's baseline step from unknown measurement noise into an internal state event that can be recognized and managed by the controller. By actively isolating and adjusting the output during baseline stabilization, the control system avoids amplifying sensor spurious signals into violent actuator actions, thus enabling the control loop to have higher operational safety when facing complex dynamic conditions. Attached Figure Description

[0024] Figure 1 This is a flowchart of the in-situ zero-point self-calibration logic of the electrochemical ammonia sensor that integrates environmental parameters according to the present invention.

[0025] Figure 2 This is a comparison chart of the sensor measurement response values ​​of the zero-point self-calibration under transient environmental humidity shock of the present invention and the control group;

[0026] Figure 3This is an event-driven state transition logic diagram of the self-calibration method of the present invention in a control system. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0028] This invention discloses an in-situ zero-point self-calibration method for an electrochemical ammonia sensor that integrates environmental parameters. It is applied to a control system including an electrochemical ammonia sensor, an environmental parameter acquisition unit, and a controller. The controller acquires the sensor's measured values ​​and the environmental temperature and humidity parameters from the environmental parameter acquisition unit. The core logic at the control system level abandons the technical path of establishing a static compensation model. Instead, the controller monitors the first derivatives of the environmental temperature and humidity parameters in real time and compares them with a preset environmental disturbance threshold to identify baseline transition risk events that may cause discontinuous step jumps in the sensor baseline. In response to this event, the controller immediately performs active intervention, maintaining its output signal for regulating the controlled object at the output value immediately before the event identification. This control action logically establishes a quasi-static technical premise for subsequent calibration processes. The controller then initiates a stabilization window, constrained by this quasi-static premise, ensuring the sensor reaches zero point again under the new and old environments before and after the stabilization window. The difference between the sensor measurements at chemical equilibrium is determined as a step change in the sensor zero-point baseline due to environmental transients. The controller then accumulates this step change into the cumulative zero-point offset state variable. After the stabilization window ends, the maintenance of the control output signal is released, and closed-loop control of the controlled object is restored based on the calibrated sensor signal. In the actual operation of the control system, the measurement baseline (zero point) of the electrochemical sensor will experience a nonlinear step change in response to drastic and transient changes in ambient temperature, especially humidity. This step change is not caused by the process variable being measured (such as ammonia concentration). Traditional control systems cannot distinguish between such sensor spurious signals and real process disturbances, thus making incorrect adjustment actions and causing instability in the main control flow. To address this challenge, the controller in this invention monitors the transient characteristics of environmental parameters. The controller performs high-frequency oversampling of ambient temperature and humidity parameters with a sampling period shorter than the main control cycle (e.g., 100 milliseconds), to monitor the environmental parameters. , The signal is acquired, and its first derivative (rate of change) is calculated in real time through continuous differential operation or digital filtering algorithm. and The controller will determine the absolute value of the first derivative. and With the preset environmental disturbance threshold ( Real-time comparisons are performed; the determination of the environmental disturbance threshold is based on the engineering calibration process, and its setting needs to balance response sensitivity and system stability. A threshold that is too low, such as below 0.1, is problematic. A threshold of 1.0 / second may cause the controller to frequently trigger false triggers in response to normal fluctuations in the process environment, while an excessively high threshold, such as above 1.0, could also lead to problems. If the rate is too low ( / second), it may miss capturing the true baseline step event. In typical industrial applications, offline testing can be performed by artificially applying different gradients of temperature and humidity shocks under target operating conditions (such as flues or farms). The minimum rate of environmental change corresponding to when the sensor baseline begins to exhibit a nonlinear step can be observed. Based on this, a certain safety margin can be added to set the threshold, for example, it can be set to [value missing]. or When the controller detects that the absolute value of any environmental change rate exceeds a certain threshold... When a baseline transition risk event is triggered, the controller determines that a baseline transition risk event has been triggered. This identification mechanism enables the control system to predict the risk of sensor signals becoming inaccurate from the source of the disturbance (environmental transients), providing a basis for subsequent active intervention.

[0029] In response to the identification of a baseline transition risk event, the direct threat to the control system is that its closed-loop regulation logic is interfered with by the sensor's spurious signal. To avoid this risk and seize the opportunity to complete self-calibration, the controller immediately executes an isolation-stabilization control procedure. The first step (step a) of this procedure is to maintain its regulation output signal to the controlled object at the output value just before the event was identified. This action is called regulation output freeze in the context of control systems. The controller retrieves and stores the last sensor reading considered reliable just before the event was identified, i.e., the first sensor measurement value, denoted as . Taking the ammonia injection actuator in a selective catalytic reduction (SCR) denitrification system as an example, the output signal is adjusted to the control signal of the ammonia injection valve. The controller temporarily cuts off the control loop's regulation of the ammonia injection quantity by maintaining this control signal, such as the PWM duty cycle signal, at the value before the event, for example, a 35% duty cycle. This freezing behavior is the core of this solution. Based on the controller's active intervention, a key quasi-static technical premise is logically established, namely, the controller assumes that during the subsequent short stabilization window (e.g., 60 seconds), since the actuator (ammonia injection valve) of the main process is frozen, the actual process variable (ammonia concentration) is basically stable and the change is negligible. In the second step (step b) of this procedure, the controller starts the stabilization window while freezing the output, and continuously monitors the sensor measurements within this window. rate of change The controller waits for the sensor's electrochemical state to change under new environmental parameters (new). , The criteria for achieving a new balance are: The signal stability threshold is lower than the preset threshold. ),Should The setting should be lower than the sensor's rate of change under normal process fluctuations but higher than its noise floor; it can be set to 0.05% / second of the sensor's full scale. Ensure this procedure will always exit; the stabilization window has a preset maximum duration, such as 60 to 90 seconds. If within the window... Before reaching its maximum duration, it was already below... If a preset stabilization time (e.g., 5 seconds) is maintained, the controller determines that the sensor has reached a new steady state ahead of schedule and ends the stabilization window early. If the rate of change has not reached the target by the end of the maximum duration, the controller forcibly reads the sensor reading at the end of the maximum duration. At the end of the stabilization window, the controller acquires the sensor reading at that moment as the second sensor measurement value, denoted as... This isolation-stability procedure provides calibrable (quasi-static) logical conditions through the active intervention of the control system (freezing the output) and waits for the measuring element to reach physical equilibrium.

[0030] In acquiring and After establishing the quasi-static technical premise of the process with two values, the controller executes the calibration update logic; in step d, the controller, constrained by the quasi-static technical premise, will... and The difference between them, i.e. Logically determined or primarily attributed to the step change in the sensor zero-point baseline due to environmental transients; step e, the controller maintains the cumulative zero-point offset state variable in its internal non-volatile memory. ), perform an update operation, and accumulate the step value into the state variable, that is To avoid this calibration action causing new step disturbances to the control loop itself, a gradual update strategy is preferred when performing step e or subsequent corrections. This strategy calculates the step amount... Then, do not apply it all at once. Instead, the step value is smoothly updated to the cumulative zero-point offset state variable over multiple control cycles, such as 10 to 50 cycles, using a preset weighted average algorithm or by incrementing in preset small steps. For example, a weighted average algorithm is used, where the updated offset is the sum of the current offset and the target offset. The weighted average of the values ​​is obtained by taking the weighting factor. The gain can be small, between 0.05 and 0.2; after the stabilization window ends, the controller releases the control signal maintaining the regulating output signal, i.e., unfreezes the ammonia injection valve, and determines the gain based on the difference between the sensor measurement value and the updated cumulative zero-point offset state variable. This restores the closed-loop regulation of the controlled object; through this event-driven attribution and update closed loop, the control system can autonomously track the sensor baseline step caused by environmental transients in situ without the need for external physical standard gas.

[0031] To improve the reliability and adaptability of the self-calibration method, the present invention may further include one or more enhancement procedures; in the implementation, the controller establishes and maintains an environmental modulation decoupling model, which describes the intrinsic modulation relationship between environmental temperature and humidity parameters and the sensor zero-point baseline, and can be used for fitting based on historical data. Lookup table or polynomial model; after step d, the controller can also use this model based on environmental parameters before and after the event ( and ), predicting the step size of the model The controller calculates the event-driven difference. The step size predicted by the model The calibration deviation between the two events is recorded and correlated with the calibration event. This deviation can be used to assess the confidence level of the event-driven calibration or as a basis for subsequent optimization. The basis of the model; in another embodiment, it is for calibrating the long-term slow drift of the sensor or for the above-mentioned The model provides online calibration data, and the controller also performs a clean window identification step. This step monitors the time-series patterns of sensor measurements and environmental parameters. When the sensor measurement is within a preset low percentile range of historical data (e.g., below the 5th percentile of historical data) and its variance is below a preset stability threshold, these two factors indicate that there is a high probability of no ammonia in the process environment, the signal is stable, and the environmental parameters are within a preset typical range (e.g., under non-extreme temperature and humidity conditions), the controller identifies the clean window. During the clean window period, the controller can acquire the sensor measurement at that moment and the corresponding environmental parameters, using this data to accumulate zero-point offset state variables. Perform slow corrections or decouple the model from the environment. Online calibration or parameter updates are performed. To ensure the accuracy of clean window identification, a confidence assessment mechanism can be introduced. The controller comprehensively evaluates the duration of the clean window, data stability, and the degree of agreement with the predictions of the environmental modulation decoupling model to generate a confidence score. Only when the confidence score exceeds a preset confidence threshold will subsequent correction or calibration actions be performed. To provide a safety net for the control system, the method can also include a dual verification and backtracking arbitration mechanism for calibration actions. After executing step e, the controller immediately enters the post-hoc monitoring period to check whether the short-term stability and response characteristics of the sensor measurements near the new baseline are within a preset reasonable range. For example, it checks whether the control output has changed. Saturation or severe oscillation; the controller records calibration events and their context data as a traceable log. If a calibration event causes abnormal data behavior during the post-monitoring period or subsequent operation, the controller initiates arbitration, automatically rolls back the accumulated zero-point offset state variable to the state before the calibration event, and marks the failed calibration event. The controller can also deploy an adaptive learning framework, which continuously absorbs all operational data, including baseline transition risk event identification records, calibration records, clean window identification records, verification deviations, and arbitration execution records. It periodically uses this data to iteratively optimize environmental disturbance thresholds, signal stability thresholds, clean window identification parameters, or internal parameters of the environmental modulation decoupling model.

[0032] Example 1: In the selective catalytic reduction (SCR) denitrification control system, the controller applies closed-loop regulation to the ammonia injection actuator, which is the controlled object, based on the sensor measurement value of the electrochemical ammonia sensor, in order to maintain the stability of the outlet NOx concentration. At a certain moment, the bypass damper upstream of the process opens, resulting in a stream of high-temperature and high-humidity flue gas. The ambient temperature and humidity parameters change transiently, impacting the sensor. In a control system that does not use the method of this invention, this environmental impact causes a step change in the sensor baseline, causing the sensor measurement value to jump from a stable value of 5 ppm to 15 ppm without changing the actual ammonia concentration. Traditional controllers cannot recognize this signal, interpreting it as a process disturbance, i.e., ammonia escape, and immediately execute incorrect regulatory actions, such as drastically shutting off the regulating output signal of the ammonia injection valve. This directly leads to insufficient actual ammonia supply in the system, excessive NOx emissions, and oscillations and instability in the entire control loop. In the control system using the method of this invention, while executing the main regulating logic, the controller simultaneously collects ambient temperature and humidity parameters at high frequency through the environmental parameter acquisition unit and calculates their first derivatives in real time. When high-temperature and high-humidity flue gas impacts the sensor, the controller monitors the absolute value of the first derivative, such as the humidity change rate. If the ambient disturbance threshold is momentarily exceeded, the controller immediately identifies a baseline transition risk event. In response to this event, the controller does not adjust using the already increased sensor measurement value of 15 ppm. Instead, it immediately executes step a, maintaining its output signal for the ammonia injection actuator (i.e., the ammonia injection valve control signal) at the value just before the event identification and storing the first sensor measurement value of 5 ppm. The controller then initiates a stabilization window, during which it continuously monitors the rate of change of the sensor measurement value until the sensor reaches electrochemical equilibrium again under the new temperature and humidity environment, and the rate of change is lower than the preset signal stabilization threshold. The controller then obtains the second sensor measurement value of 14.8 ppm at this moment. By maintaining this control action with the adjustment output signal in step a, the technical premise that the process in step c is in a quasi-static state is established. Based on this, the controller confirms that during the stabilization window, the actual ammonia concentration remains stable because the valve opening remains unchanged.

[0033] Based on the constraints of the quasi-static technical premise, the controller then executes step d, determining the difference of 9.8 ppm between the second sensor measurement value of 14.8 ppm and the first sensor measurement value of 5 ppm as a step change in the sensor zero-point baseline due to environmental transients; the controller then adds this 9.8 ppm step change to the internally maintained cumulative zero-point offset state variable. In this context, this is step e; after the stabilization window ends, the controller releases the maintenance of the ammonia injection valve's regulating output signal and, according to... The calculated difference restores the closed-loop regulation of the controlled object; at this moment, although the sensor's raw readings... It stabilized at 14.8 ppm, but the controller was used for adjustment. Value The value is consistent with the measurement value before the event occurred; the controller's adjustment output signal thus remains stable, avoiding erroneous valve closing adjustments, NOx emissions are always kept within the acceptable range, the control system autonomously completes in-situ tracking of the new baseline of the sensor, and maintains the smooth operation of the main control process.

[0034] Example 2: To objectively verify the effectiveness of the method of the present invention at the control system level in suppressing sensor baseline jumps caused by transient changes in environmental parameters and ensuring the smooth operation of the main control process, a closed-loop control test platform was built. The test platform includes a controller for simulating the SCR denitrification process, which has a 1-second main control cycle; an electrochemical ammonia sensor (range 0-50ppm) installed in a temperature- and humidity-controlled environmental chamber to provide sensor measurements; an environmental parameter acquisition unit that monitors the temperature and humidity parameters in the environmental chamber with a sampling period of 100 milliseconds; a process simulator that provides the sensor with a constant standard gas of 5.0ppm representing the actual ammonia concentration; and the controller's adjustment output signal (simulating the ammonia injection valve control signal, 0-100%) sent to a recorder. Two test groups were set up: one group was a control group, where the controller used standard PID control logic and directly used the sensor measurements. As a process variable (PV), it is compared with a setpoint (SP) of 5.0 ppm, and the adjusted output signal is calculated and output; the other group is the experimental group of this invention, in which the controller uses the same PID control logic and setpoint, but its PID process variable input is a calibrated value. The controller fully deploys the self-calibration method of this invention, including real-time calculation of the first derivative of environmental parameters, identification of baseline transition risk events, execution of isolation-stabilization control procedures, and updating of cumulative zero-point offset state variables. The key parameters for the experimental group of this invention are set as follows: environmental disturbance threshold. Set as Maximum duration of the stability-seeking window Set to 60 seconds, signal stability threshold The setting was 0.02 ppm / second; the experimental procedure was as follows: at 0 seconds, both sets of experiments were started, and the environmental chamber was kept in a stable environment of 25°C and 40%RH. The process simulator continuously introduced 5.0 ppm of standard ammonia gas; after the system had been running stably for 10 seconds, at 11 seconds, a stream of water vapor was rapidly injected into the environmental chamber, causing the ambient humidity to jump rapidly from 40%RH to 75%RH within 2 seconds, and then remain stable; the key data of the two sets of experiments were continuously monitored and recorded from 0 seconds to 80 seconds, and the typical data obtained are shown in Table 1.

[0035] Table 1: Comparison of Control System Responses under Environmental Transient Shocks

[0036]

[0037] Analysis of the data in Table 1 shows that the ambient humidity began to change drastically at 11 seconds. The rate reached 15.3%RH / second, exceeding the threshold of 1.0%RH / second; the control group's controller, due to direct use... At 11 seconds, the sensor measurement value The temperature jumped from 5.02 ppm, and the controller misinterpreted this as a significant deviation of the process ammonia concentration (PV) from the setpoint (SP). Its regulating output signal immediately reacted drastically, rapidly dropping from 35.3% to 0.0% at 12 seconds, effectively closing the control valve completely. This erroneous control action would lead to an interruption of ammonia supply and excessive NOx emissions under real-world conditions. In the experimental group of this invention, the controller also detected this at 11 seconds. If the environmental disturbance threshold is exceeded, a baseline transition risk event is immediately identified, and an isolation-stabilization control procedure is executed. The output signal is adjusted to maintain 35.3% of the value at 10 seconds, while the first sensor measurement value of 5.02 ppm is stored immediately before the event identification. During the stabilization window from 11 seconds to 71 seconds, although... The reading has jumped to around 14.9 ppm. The adjusted output signal of the experimental group of this invention remains constant at 35.3%, effectively isolating the control loop from transient spurious signals from the sensor. At the end of the 71-second, 60-second stabilization window, the controller acquires the second sensor measurement at this moment, which is 14.95 ppm. Based on the quasi-static premise, the difference between the second sensor measurement and the first sensor measurement, i.e., the step value, is calculated. ppm, updated immediately The value is 9.93 ppm; starting from 72 seconds, the controller releases the hold state and resumes closed-loop regulation. At this time, the process variable used for PID calculation is... The value of ppm is basically consistent with the set value of 5.0ppm, and the adjusted output signal remains stable around 35.3%. The test data shows that the method of the present invention can autonomously identify risks when encountering severe environmental parameter transients, establish quasi-static calibration premise through active intervention (maintaining the adjusted output), and autonomously complete the tracking and calibration of the sensor zero-point baseline step in situ.

[0038] Example 3: This example combines Figures 1 to 3 The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters is explained, such as... Figure 1As shown, the controller performs the steps of acquiring sensor measurements and environmental parameters. The sensor measurements come from an electrochemical ammonia sensor, and the environmental parameters include temperature and humidity from the environmental parameter acquisition unit. The controller calculates the first derivative of the environmental parameters, which involves high-frequency oversampling and differential or filtering operations. The controller then determines whether the absolute value of the first derivative is greater than a preset environmental disturbance threshold to identify baseline transition risk events. If the determination is negative, monitoring continues; if the determination is positive, an event is triggered, and the controller performs active intervention step a, i.e., maintaining the regulation output signal frozen and storing the first sensor measurement value, and simultaneously initiating the stabilization window step b. Within this window, the rate of change is monitored until it is less than the signal stabilization threshold, at which point the second sensor measurement value is acquired. After stabilization is completed, the controller performs active intervention step a, i.e., maintaining the regulation output signal frozen and storing the first sensor measurement value, and simultaneously initiating the stabilization window step b. Within this window, the rate of change is monitored until it is less than the signal stabilization threshold, at which point the second sensor measurement value is acquired. In step d, the controller determines the baseline step amount. Based on the quasi-static premise of the process, the step amount = second value - first value. Then, step e updates the cumulative zero-point offset by adding the step amount to the cumulative zero-point offset state variable. This step adopts a progressive update strategy. The controller resumes closed-loop regulation, releases the regulation output maintenance, and uses the measured value and cumulative zero-point offset for adjustment to bring the system back to normal operation. This process can also interact with the cleanroom window identification process, which is used to identify the zero-concentration background moment. Its results are used to correct / calibrate the cumulative zero-point offset or the environmental modulation decoupling model. The step amount determined in step d can also be used for verification and recording, i.e., inputting into the environmental modulation decoupling model. This model predicts the model step amount based on environmental parameters, calculates the verification deviation, and records it.

[0039] like Figure 2 As shown, the horizontal axis represents time (s), the left vertical axis represents the sensor measurement value (ppm), and the right vertical axis represents the ambient humidity (%RH). The ambient humidity curve in the graph shows that the environment changes drastically at 11 seconds, with the humidity jumping from 40%RH to 75%RH. The sensor measurement curve shows that during this humidity jump, the measured value incorrectly jumps from 5ppm to 15ppm. However, the actual sensor measurement curve shows that despite the same humidity shock, the measured value remains stable at 5ppm without a baseline step. Figure 3As shown, this logic includes three core states: the normal closed-loop adjustment state, representing normal system operation; the isolation stabilization state, representing frozen output, waiting for sensor stabilization; and the post-calibration monitoring period state, representing calibration execution and effect verification. The system starts from the normal closed-loop adjustment state. When a baseline transition risk event is identified and the derivative of the environmental parameter is greater than the threshold, a transition to the isolation stabilization state is triggered. When the stabilization window ends and the signal stabilizes or times out, a transition to the post-calibration monitoring period is triggered. During the post-calibration monitoring period, if the monitoring period ends and the response is reasonable, it is considered successful, the system resumes adjustment and returns to the normal closed-loop adjustment state. If abnormal data behavior is detected, it is considered a failure, the system performs a rollback, resumes adjustment and returns to the normal closed-loop adjustment state. In the normal closed-loop adjustment state, if a clean window is identified, an internal action of online correction will also be triggered to maintain the normal closed-loop adjustment state.

[0040] Example 4: In addition to providing event-driven step tracking, the method of the present invention also includes the establishment of an environmental modulation decoupling model and an online calibration procedure based on cleanroom window identification to solve the problem of long-term, continuous zero-point baseline drift caused by slow changes in environmental parameters or sensor aging. The initial state definition procedure of the environmental modulation decoupling model in the controller is as follows: Before sensor deployment or during a predetermined offline maintenance window, the electrochemical ammonia sensor and the environmental parameter acquisition unit are placed in an environmental chamber with precise temperature and humidity control, and zero-concentration standard ammonia gas, i.e., 0 ppm, is introduced into the environmental chamber. The controller executes an automated parameter scanning program within the sensor's typical operating range, from 10... Up to 40 (with 5) Environmental parameters were set point by point, with humidity ranging from 30%RH to 80%RH (in 10%RH increments); at each ( At the set point, the controller waits until both the sensor signal and environmental parameters have stabilized before recording the sensor measurement value at that moment. Since the introduced gas is known to have zero concentration, this The value is defined as being in that specific ( The physical baseline value under the condition; the controller will include all ( Data points are stored, and an initial environmental modulation decoupling model is generated through polynomial fitting or interpolation algorithms. The model is stored in the controller's non-volatile memory for subsequent online operation. After the control system is deployed in the control field, the controller performs event-driven calibration while simultaneously initiating the cleanroom window identification step. The controller monitors the time-series patterns, variance fluctuations, and environmental parameters of sensor measurements in real time. Internally, the controller maintains a long-term statistical histogram of sensor measurements based on the past 720 hours of operating data and calculates its historical low percentile interval, taking the 5th percentile as an example. When the controller detects the current sensor measurement... For a sustained period of time, such as 1800 seconds, the value is below the 5th percentile, and the short-term fluctuation variance of the sensor measurement is below the preset stability threshold (indicating stable signal and no ammonia events), and environmental parameters ( When the target gas concentration is also in a stable state (i.e., no baseline transition risk event is triggered), the controller identifies the clean window; this procedure enables the control system to autonomously infer the background moment when the target gas concentration logically approaches zero in a real process environment where physical standard gases are not available.

[0041] Upon identifying the clean window, the controller immediately executes a confidence assessment and online calibration procedure; the controller collects multiple sets of data points during the clean window period, and for each data point ( The controller modulates the decoupling model based on the stored environment. Calculate the model prediction baseline value under this environment. The controller then calculates the current total zero-point offset. ,in This is the step offset accumulated from event-driven calibration; the controller calculates the sensor measurements at the clean window. Total zero offset residuals between ,this The value represents the result of factors such as long-term aging of the sensor that were not included in the model. and step quantity This can explain slower drift; the controller's response to this... The confidence level of the amplitude and stability is assessed when the It remains stable within a small range, such as for 30 consecutive minutes. The mean is at If the value exceeds a preset confidence threshold within the ppm range, the controller confirms this. To represent the true long-term drift; the controller performs online calibration, employing a gradual update strategy, and then... The value or its weighted average is smoothly corrected to the cumulative zero-point offset state variable in small steps. In this, using this ( The data points are used as new training samples to decouple the environment modulation model. The internal parameters are updated online iteratively; this procedure works in conjunction with the event-driven procedure to enable the control system not only to cope with baseline steps caused by environmental transients, but also to autonomously correct the slow drift caused by long-term sensor aging.

[0042] Example 5: When the control system of the present invention is initially deployed in a specific control application site, the preset parameters used for cleanroom window identification can be determined through standardized field engineering calibration procedures. Before engaging closed-loop regulation, the controller enters an initial data acquisition mode for 48 hours. During this period, the controller only collects and stores the sensor measurements from the electrochemical ammonia sensor and the ambient temperature and humidity parameters from the environmental parameter acquisition unit at high frequency, without performing any adjustment output or self-calibration actions. After this 48-hour initial state definition phase, the controller performs statistical analysis on all collected sensor measurement data, calculates its cumulative probability distribution, and sets the 5th percentile value of the data sequence, for example, 2.1 ppm, as the initial upper limit of the preset historical data statistical low percentile interval. The controller calculates that all fluctuation variances in this data sequence are below a specific benchmark, for example, 0.1. The average variance of the stable data segment, for example, 0.03. The preset stable threshold is set as the clean window identification step; the preset typical range of environmental parameters is determined based on the 10% and 90% quantiles of ambient temperature and humidity in this 48-hour data. This procedure ensures that the self-calibration logic of the control system is anchored to the actual working condition baseline of the current site from the beginning of operation.

[0043] The adaptive learning framework deployed by the controller iteratively optimizes the environmental disturbance threshold and signal stability threshold through a closed-loop adjustment process based on feedback from historical calibration events. Internally, the controller maintains an execution log for the dual verification and backtracking arbitration mechanism for calibration actions, recording the triggering context of each baseline transition risk event. , The adaptive learning framework periodically analyzes this log, for example, every 24 hours. If the framework finds that the trigger rate of the backtracking arbitration mechanism exceeds the upper limit, the controller determines that the current environmental disturbance threshold or signal stability threshold setting is too aggressive, leading to incorrect calibration of non-steady states. The controller then increases the environmental disturbance threshold or signal stability threshold by a preset small step (e.g., increasing the original threshold by 5%). If the framework finds that the verification deviation of the calibration action, i.e., the deviation between the difference and the model step, is consistently and stably less than the lower limit, for example, 0.1 ppm, the controller determines that the current threshold setting is too conservative and may miss effective calibration opportunities. The controller then appropriately lowers the threshold by a preset small step. This adaptive adjustment mechanism enables the calibration strategy of the control system to autonomously find and maintain a dynamic balance between sensitivity and stability.

[0044] Example 6: Environmental disturbance threshold for identifying baseline transition risk events in the method of the present invention. And the threshold for determining signal stability using a stability-finding window. Its initial settings in the controller can be determined through standardized engineering calibration procedures; this is for calibrating environmental disturbance thresholds. The electrochemical ammonia sensor and environmental parameter acquisition unit were placed in a temperature- and humidity-controlled environmental chamber. Zero-concentration standard ammonia gas was introduced into the chamber, and the sensor readings were collected. Once stabilized, this stable value will be used as the benchmark. The controller adjusts the humidity at a series of increasing rates of change. Beginning, with Incrementing step size to A transient humidity shock was applied to the environmental chamber, and after each shock and subsequent restabilization, the new stable values ​​of the sensor measurements were recorded. Compared with the benchmark The difference between them, i.e. ; Controller drawing Follow The system reads the changing response curve and automatically identifies the inflection point where the slope of the curve first shows a non-linear increase. It then identifies the corresponding inflection point. Values ​​such as An additional safety margin of 20% is added to determine the environmental disturbance threshold. for For the rate of temperature change The calibration follows the same procedure; to determine the stability threshold of the calibration signal. The controller, under the condition that zero-concentration standard ammonia gas is introduced into the environmental chamber, continuously collects sensor measurements for 10 minutes at a high frequency, such as 100 milliseconds. Time series, calculate the first difference of this time series ( The standard deviation of the noise level is defined as the noise floor level of the sensor. For example, 0.008 ppm / second; under the condition of introducing a medium concentration of standard ammonia gas, such as 10 ppm, and superimposed with a low-frequency sinusoidal disturbance signal representing typical process fluctuations, the controller collects data again for 10 minutes, calculates the standard deviation of the rate of change of the sensor measurements under this condition, and defines it as the normal process fluctuation level. For example, 0.05 ppm / second; signal stability threshold Set to a value between the two, so as to For example, that is This procedure ensures that the noise level is both above the background noise level and below the normal process fluctuation.

[0045] The calibration action employs a dual verification and backtracking arbitration mechanism, with the following specific control logic: The controller executes step e, which involves accumulating the step value to... Immediately afterwards, a fixed-duration post-test monitoring period is initiated, such as 300 seconds. During this monitoring period, the controller monitors two core control system indicators. The first indicator is short-term stability, and the controller monitors the calibrated measured values. (Right now Whether a persistent, non-convergent oscillation occurs near a set value is quantitatively determined based on the following criteria: Does the variance during the post-monitoring period exceed the normal process fluctuation level? The first indicator is the preset multiple (e.g., 5 times); the second indicator is the regulating output characteristic. The controller monitors whether its own regulating output signal continuously touches its 0% or 100% saturation limit during the posterior monitoring period. The quantitative judgment is based on whether the cumulative time of the regulating output signal being in the saturation limit state exceeds 20% of the total duration of the posterior monitoring period. If the above indicators are triggered, the controller determines that this calibration event has caused abnormal data behavior of the control system, immediately initiates arbitration, and automatically adjusts the cumulative zero-point offset state variable. Roll back to the state value before the calibration event occurred, and record the failed calibration event and its context data in the traceable log.

[0046] 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.

[0047] 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 in-situ zero-point self-calibration of an electrochemical ammonia sensor integrating environmental parameters, applied to a control system including an electrochemical ammonia sensor, an environmental parameter acquisition unit, and a controller, characterized in that, The methods include: The controller acquires the sensor measurements from the electrochemical ammonia sensor, as well as the ambient temperature and humidity parameters from the environmental parameter acquisition unit; the controller calculates the first derivatives of the ambient temperature and humidity parameters; the controller compares the absolute value of the first derivative with a preset environmental disturbance threshold to identify baseline transition risk events. In response to the identification of a baseline transition risk event, the controller executes the following steps: Step a, maintaining its adjustment output signal for the controlled object at the output value just before the event identification, and storing the first sensor measurement value just before the event identification; Step b, starting a stabilization window, monitoring the sensor measurement values ​​within this window until the rate of change of the sensor measurement values ​​is lower than a preset signal stabilization threshold, then acquiring the second sensor measurement value at this moment, and the controller then executes the following steps: Step c, based on the fact that the adjustment output signal has been maintained at the output value in step a, establishing a quasi-static technical premise for the process of the controlled object from the moment before the event identification to the moment of acquiring the second sensor measurement value; Step d, constrained by the quasi-static technical premise, determining the difference between the second sensor measurement value and the first sensor measurement value as the step amount of the sensor zero-point baseline caused by environmental transients; Step e, accumulating the step amount into the cumulative zero-point offset state variable; After the stabilization window ends, the controller releases the maintenance of the regulation output signal and resumes closed-loop regulation of the controlled object based on the difference between the sensor measurement value and the updated cumulative zero-point offset state variable.

2. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 1, characterized in that, In step b, the stabilization window has a preset maximum duration; If, within the stabilization window, the rate of change of the sensor measurement value is lower than the preset signal stabilization threshold and remains stable for a preset time before reaching the maximum duration, the controller will end the stabilization window early and acquire the second sensor measurement value; if the rate of change is still not lower than the preset signal stabilization threshold until the end of the maximum duration, the controller will read the sensor reading at the end of the maximum duration as the second sensor measurement value.

3. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 1, characterized in that, The first derivative is obtained in real time by the controller through high-frequency oversampling of ambient temperature and humidity parameters, and by continuous difference operation or digital filtering algorithm.

4. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 1, characterized in that, The method also includes: the controller establishing and maintaining an environmental modulation decoupling model, which describes the intrinsic modulation relationship between environmental temperature and humidity parameters on the sensor zero-point baseline; after step d, the controller also performs: using the environmental modulation decoupling model, predicting the model step based on the environmental parameters at the moment before event identification and the moment when the second sensor measurement value is obtained; calculating the verification deviation between the difference and the model step, and recording the verification deviation in association with the calibration event of step e.

5. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor integrating environmental parameters according to claim 4, characterized in that, The method further includes: the controller also performs a clean window identification step, which includes: monitoring the time series pattern of sensor measurements and environmental parameters; identifying a clean window when the sensor measurements are within a preset historical data statistical low percentile interval and their fluctuation variance is lower than a preset stability threshold and the environmental parameters are within a preset typical range; during the clean window period, the controller acquires the sensor measurements and the corresponding environmental temperature and humidity parameters, and uses the sensor measurements and the corresponding environmental temperature and humidity parameters to correct the cumulative zero-point offset state variable or to calibrate the environmental modulation decoupling model online.

6. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 5, characterized in that, The clean window identification process also includes: the controller introducing a confidence assessment mechanism; the controller comprehensively assesses the duration of the clean window, the data stability of the sensor measurements, and the degree of agreement with the predictions of the environmental modulation decoupling model to generate a confidence score; only when the confidence score exceeds the preset confidence threshold will the controller perform correction of the cumulative zero-point offset state variable or online calibration of the environmental modulation decoupling model.

7. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 1, characterized in that, When executing step e or performing correction, the controller adopts a progressive update strategy. The progressive update strategy includes: after calculating the step amount or correction amount, the controller does not apply it to the cumulative zero-point offset state variable all at once, but smoothly updates the step amount or correction amount to the cumulative zero-point offset state variable over multiple control cycles using a preset weighted average algorithm or by increasing it in preset small steps.

8. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor integrating environmental parameters according to claim 1, characterized in that, The method also includes a dual verification and retrospective arbitration mechanism for calibration actions. The mechanism includes: after executing step e, the controller immediately enters the post-monitoring period to check whether the short-term stability and response characteristics of the sensor measurements near the new baseline are within a preset reasonable range; the controller records the calibration event and its context data of step e as a traceable log; if the calibration event causes abnormal data behavior during the post-monitoring period or in subsequent operation, the controller initiates arbitration and automatically rolls back the accumulated zero-point offset state variable to the state before the calibration event.

9. The in-situ zero-point self-calibration method for an electrochemical ammonia gas sensor incorporating environmental parameters according to claim 1, characterized in that, The controlled object is the ammonia injection actuator in the selective catalytic reduction (SCR) denitrification system, and the output signal is the control signal of the ammonia injection valve. The controller maintains the control signal of the ammonia injection valve at the output value just before the event recognition through step a, so as to avoid incorrect adjustment of the control signal of the ammonia injection valve due to the step of the sensor zero-point baseline during the stabilization window.

Images