Isotope gas concentration intelligent regulation method and system

Abstract

The application provides an isotope gas concentration intelligent regulation method and system. The method comprises the following steps: constructing a historical regulation state according to historical concentration sampling values and historical control action values. The historical regulation state comprises a concentration main level and an action residual intensity. Calculating a delay release amount according to the concentration main level and the action residual intensity. Calculating a plurality of concentration prediction values according to a current concentration sampling value, the delay release amount, the concentration main level and the action residual intensity. All the concentration prediction values constitute a future prediction concentration sequence. Obtaining a target concentration set value of a current period. Calculating a control demand amount of the current period according to the future prediction concentration sequence, the target concentration set value, a time sequence attenuation coefficient and an extreme value risk coefficient. Calculating a target control instruction of the current period according to the control demand amount, a basic regulation coefficient and an absolute value of the control demand amount. Regulating the isotope gas concentration according to the target control instruction. The accuracy of the isotope gas concentration regulation is improved.

Description

Technical Field

[0001] This invention belongs to the field of isotope gas concentration regulation, and particularly relates to an intelligent method and system for isotope gas concentration regulation. Background Technology

[0002] In precision experiments related to stable isotope gas applications, agricultural breeding, plant photosynthetic metabolism research, medical metabolic tracing, and small-scale closed culture systems, it is typically necessary to maintain continuous, stable, and high-precision control of the concentration of a specific isotope gas within a confined space. Compared to conventional gas control scenarios, these applications not only demand higher precision in concentration maintenance but also face significant challenges. Due to the high cost of isotope gases, the long duration of experiments, and the limited volume and closed boundaries of the chamber, overshoot, pullback, oscillation, or repeated pumping during control can directly increase gas consumption, disrupt the consistency of experimental conditions, and affect the accuracy and comparability of results. Related technologies utilize real-time concentration measurements as the primary control basis, employing valve switching, pumping device activation / deactivation, or conventional closed-loop regulation to perform gas replenishment or pumping operations only after the concentration deviates from the target value. While these methods can achieve basic regulation in general ventilation or ordinary concentration maintenance scenarios, they often struggle to balance accuracy, control stability, and gas utilization efficiency when used for precise control of isotopic gases in confined spaces. The fundamental reason is that changes in gas concentration within the chamber are not an immediate reflection of control actions. Instead, they are influenced by factors such as local gas injection, mixing and diffusion, spatial transmission, and residual effects, exhibiting significant delayed manifestation and continuous release characteristics. In other words, after a control action is initiated, the concentration value measured by the sensor at that moment is only the already manifested result, while control actions from previous cycles often continue to affect concentration evolution in subsequent periods. If the control method only corrects based on the current measured value, it is easy to misjudge historical effects that have not yet been fully released as new sources of deviation, leading to problems such as overcompensation, frequent reverse adjustments, and increased steady-state fluctuations. Therefore, improving the accuracy of isotopic gas concentration regulation has become an urgent technical problem to be solved. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent method and system for controlling isotope gas concentration, which can improve the accuracy of isotope gas concentration control.

[0004] To achieve the above objectives, a first aspect of the present invention provides a method for intelligent control of isotope gas concentration, the method comprising:

[0005] Multiple historical concentration sampling values ​​and multiple historical control action values ​​of a historical window length are obtained, and a historical regulation state is constructed based on the historical concentration sampling values ​​and the historical control action values; wherein, the historical regulation state includes the main concentration level and the residual intensity of the action;

[0006] The delayed release amount is calculated based on the main concentration level and the residual intensity of the action, and the current concentration sample value is obtained. Multiple concentration prediction values ​​are calculated based on the current concentration sample value, the delayed release amount, the main concentration level, and the residual intensity of the action; wherein, all concentration prediction values ​​constitute the future predicted concentration sequence.

[0007] Obtain the target concentration setpoint for the current period, and calculate the control demand for the current period based on the predicted future concentration sequence, the target concentration setpoint, the preset time-series decay coefficient, and the preset extreme value risk coefficient.

[0008] The target control command for the current period is calculated based on the control demand, the preset basic adjustment coefficient, and the absolute value of the control demand. The isotope gas concentration is then adjusted according to the target control command.

[0009] Further, the step of adjusting the isotope gas concentration according to the target control command includes:

[0010] Retrieve the actual execution control instructions from the previous cycle;

[0011] The actual execution control command for the current cycle is calculated based on the actual execution control command of the previous cycle, the preset execution dynamic coefficient, and the target control command.

[0012] The isotope gas concentration is adjusted according to the actual execution control command of the current cycle.

[0013] Further, the step of calculating the actual execution control instruction for the current cycle based on the actual execution control instruction of the previous cycle, the preset execution dynamic coefficient, and the target control instruction includes:

[0014] The difference between the target control instruction and the actual execution control instruction of the previous cycle is multiplied by the execution dynamic coefficient, and then added to the actual execution control instruction of the previous cycle to obtain the actual execution control instruction of the current cycle.

[0015] Further, the step of constructing the historical control state based on the historical concentration sampling values ​​and the historical control action values ​​includes:

[0016] The historical concentration sampling value and the historical control action value at each moment are concatenated and then multiplied by the corresponding memory decay coefficient to obtain multiple first data.

[0017] Add all the first data together to get the second data, and add all the memory decay coefficients together to get the third data;

[0018] The historical control state is obtained by calculating the ratio between the second data and the third data.

[0019] Further, the step of calculating multiple concentration prediction values ​​based on the current concentration sampling value, the delayed release amount, the main concentration level, and the residual intensity of the action includes:

[0020] Get forward The concentration prediction value at each time point; wherein, the concentration prediction value at the previous time point is initially the current concentration sample value;

[0021] Add the residual intensity of the action to the delayed release amount, and then multiply by the corresponding attenuation coefficient to obtain the fourth data. Subtract the concentration prediction value of the previous moment from the main concentration level, and then multiply by the preset regression coefficient to obtain the fifth data.

[0022] Add the predicted concentration value from the previous moment, the fourth data, and the fifth data to obtain the forward [value]. The predicted concentration value at each time point.

[0023] Further, the step of calculating the control demand for the current period based on the predicted future concentration sequence, the target concentration setpoint, the preset time-series decay coefficient, and the preset extreme value risk coefficient includes:

[0024] Subtract the predicted concentration value from the target concentration setpoint at each time point, and then multiply by the corresponding time-series decay coefficient to obtain multiple sixth data points. Add all the sixth data points together to obtain the seventh data point.

[0025] The largest data is selected from all the sixth data to obtain the eighth data. The eighth data is multiplied by the extreme value risk coefficient and then added to the seventh data to obtain the control demand for the current period.

[0026] Further, the step of calculating the target control command for the current period based on the control demand, the preset basic adjustment coefficient, and the absolute value of the control demand includes:

[0027] Add the preset number to the absolute value of the control demand to obtain the ninth data. Divide the control demand by the ninth data and multiply by the opposite of the basic adjustment coefficient to obtain the target control command.

[0028] A second aspect of the present invention provides an intelligent isotope gas concentration control system, the system comprising:

[0029] An acquisition unit is used to acquire multiple historical concentration sampling values ​​and multiple historical control action values ​​for a historical window length, and to construct a historical control state based on the historical concentration sampling values ​​and the historical control action values; wherein, the historical control state includes a main concentration level and residual action intensity;

[0030] The first calculation unit is used to calculate the delayed release amount based on the main concentration level and the residual intensity of the action, obtain the current concentration sample value, and calculate multiple concentration prediction values ​​based on the current concentration sample value, the delayed release amount, the main concentration level, and the residual intensity of the action; wherein, all concentration prediction values ​​constitute the future predicted concentration sequence;

[0031] The second calculation unit is used to obtain the target concentration setpoint for the current period, and calculate the control demand for the current period based on the future predicted concentration sequence, the target concentration setpoint, the preset time decay coefficient, and the preset extreme value risk coefficient.

[0032] The control unit is used to calculate the target control command for the current period based on the control demand, the preset basic adjustment coefficient and the absolute value of the control demand, and to adjust the isotope gas concentration according to the target control command.

[0033] In a third aspect of the invention, an electronic device is provided, the electronic device including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the method described in the first aspect above.

[0034] In a fourth aspect of the invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0035] The beneficial technical effects of the present invention are at least as follows:

[0036] To address the aforementioned problems, this invention provides an intelligent method and system for isotopic gas concentration control. Its core lies in moving away from using the current concentration sampling value in isolation. Instead, it organizes recent concentration changes and recent control actions sequentially into a historical control state, enabling the system to incorporate the control effects that have occurred but not yet fully manifested into subsequent calculations. Based on this, it further extracts the current main concentration level and residual intensity of the actions from the historical control state, generating a predicted concentration sequence for several future moments. This prediction reflects the actual evolutionary characteristics of delayed release and gradual slowing of control actions within the sealed chamber. Subsequently, it constructs the control requirements for the current cycle around the entire future predicted trajectory rather than a single predicted point, incorporating future deviation distribution and extreme value risks into the control command solution. This allows the system to determine the direction and intensity of adjustment before the concentration actually exceeds the limit. Finally, the control commands are mapped to the continuous actions of valves and extraction devices through the execution dynamic process, ensuring a smooth temporal unfolding of the gas replenishment and extraction processes, thereby coordinating the control results with the gas transfer and mixing processes within the chamber. Through the above processing, this invention can continuously connect the historical control process, future concentration evolution, and current execution actions, so that the control logic no longer stays in the traditional way of "correcting the deviation after detection", but transforms into a forward-looking control method of "predicting the future trajectory based on the historical state, and then generating and executing the current action according to the future trajectory". This more effectively solves the problems of response lag, control oscillation, local overshoot, and gas waste in the control of isotope gas concentration in confined spaces, and makes the entire control process more stable, economical, and in line with actual experimental scenarios. Attached Figure Description

[0037] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0038] Figure 1 This is a flowchart of an intelligent control method for isotope gas concentration provided in an embodiment of this application.

[0039] Figure 2 This is a schematic diagram of the structure of an intelligent isotope gas concentration control system provided in an embodiment of this application. Detailed Implementation

[0040] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] Please refer to Figure 1 , Figure 1 This is a flowchart of an intelligent isotope gas concentration control method provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S101 to S104.

[0042] Step S101: Obtain multiple historical concentration sampling values ​​and multiple historical control action values ​​for a historical window length, and construct a historical regulation state based on the historical concentration sampling values ​​and historical control action values; wherein, the historical regulation state includes the main concentration level and the residual intensity of the action;

[0043] Step S102: Calculate the delayed release amount based on the main concentration level and residual intensity of the action, obtain the current concentration sampling value, and calculate multiple concentration prediction values ​​based on the current concentration sampling value, delayed release amount, main concentration level, and residual intensity of the action; wherein, all concentration prediction values ​​constitute a future predicted concentration sequence;

[0044] Step S103: Obtain the target concentration setpoint for the current period, and calculate the control demand for the current period based on the future predicted concentration sequence, the target concentration setpoint, the preset time-series decay coefficient, and the preset extreme value risk coefficient.

[0045] Step S104: Calculate the target control command for the current cycle based on the control demand, the preset basic adjustment coefficient, and the absolute value of the control demand, and adjust the isotope gas concentration according to the target control command.

[0046] In step S101 of some embodiments, the concentration change process and control action process over a period of time prior to the current moment are compiled into a unified historical control state to characterize the evolution trajectory of isotopic gas within the chamber during recent control processes. In actual implementation, a gas concentration sensor is arranged in the main mixing zone inside the chamber, and the sensor outputs continuous concentration readings at a fixed interval. Simultaneously, the lower-level controller or upper-level control program records the valve actions and pumping device actions under the same interval. To ensure that concentration changes and control actions correspond point-to-point, the controller recording interval is consistent with the concentration sampling interval. If the controller performs multiple actions within a certain sampling period, the actual action level within that period is combined into a single period action value. The action value here uses a signed unified encoding: the direction of gas replenishment is recorded as a positive value, the direction of gas extraction is recorded as a negative value, and no action is recorded as a zero value. If both gas replenishment and gas extraction occur within the same sampling period, the two types of actions are converted into a net action value under the same control scale according to the equivalent calibration relationship obtained during the system debugging phase. After this arrangement, each historical moment corresponds to both a concentration value and a directional action value, and the state calculated subsequently can simultaneously include the "concentration result" and the "regulatory process that led to the result".

[0047] After obtaining multiple historical concentration sample values ​​and multiple historical control action values ​​within a historical window length, since the original amplitude ranges of the concentration sample value sequence and the control action value sequence are usually different, the two types of sequences are first normalized to bring them into a unified calculation scale. The normalized historical concentration sample value sequence is denoted as... The normalized sequence of historical control action values ​​is denoted as .in, The value of is between zero and one. The value ranges from negative one to positive one. Specifically, a fixed-length historical window is extracted from the current moment. The length of this historical window is determined based on the duration of the control effect within the control chamber, ensuring that the window covers the recent major delayed manifestation processes. Two fixed-length circular buffers are set up in the control host: one stores several recent consecutive historical concentration sampling values, and the other stores historical control action values ​​from the same time period. Each time a new sample arrives, the latest concentration sampling value and control action value are written to the end of the buffer, while the earliest historical data is popped, thus continuously maintaining a historical window of constant length. The construction of the historical control state adopts a normalized discrete decay memory form, assigning higher weights to historical segments closer to the current moment and lower weights to earlier historical segments. Concentration segments and action segments from the same historical moment are concatenated before participating in a weighted average. That is, the historical concentration sampling value and historical control action value at each moment are concatenated and multiplied by the corresponding memory decay coefficient to obtain multiple first data points; all first data points are added together to obtain second data points, and all memory decay coefficients are added together to obtain third data points; the ratio of the second and third data points is calculated to obtain the historical control state. As shown in the following formula:

[0048] ;

[0049] in, This represents the historical control status at the current moment, calculated by the control host within the current cycle; This indicates the length of the history window, corresponding to the number of historical segments stored in the circular buffer. This represents the memory decay coefficient, which takes a value between zero and one, and is used to control the rate at which the influence of history diminishes as it is traced back over time. Indicates the first The normalized concentration sampling values ​​at each historical moment are obtained by collecting gas concentration sensors and then normalizing them. Indicates the first The normalized control action values ​​at each historical moment are extracted from the controller log, and obtained by normalization after being coded in a signed uniform manner. This means concatenating concentration sampling values ​​and control action values ​​from the same historical moment into a joint segment. The expression is derived from a discrete weighted summation, with a normalized denominator added to the summation result to ensure that the historical control state always remains within the same computational space as the input segment. The resulting state vector has its first component always between zero and one, and its second component always between negative one and positive one, facilitating subsequent steps to directly interpret it as the main concentration level and the residual intensity of the action.

[0050] A specific operating condition can provide a more intuitive understanding of how this state is formed. Suppose we take four sampling segments backward from the current time. The normalized concentration sampling value sequence is 0.22, 0.35, 0.47, and 0.55, respectively. The normalized control action value sequence is 0.00, 0.20, 0.60, and 0.60, respectively. The history window length is 4. The memory decay coefficient is 0.5, and the weight of the nearest historical segment is... , , , The weights are 1, 0.5, 0.25, and 0.125 respectively, with a total weight of 1.875. The weighted sum of the concentration components is 0.55 + 0.5 × 0.47 + 0.25 × 0.35 + 0.125 × 0.22 = 0.90, which, when divided by 1.875, yields 0.48; the weighted sum of the action components is 0.60 + 0.5 × 0.60 + 0.25 × 0.20 + 0.125 × 0.00 = 0.95, which, when divided by 1.875, yields approximately 0.51. Therefore, the current historical control state can be written as... This result indicates that the recent main concentration level is in the upper-middle range, and the residual effect of recent gas replenishment is still relatively strong. Therefore, subsequent steps will require further analysis. Afterwards, its first component can be directly interpreted as the current main concentration level, and its second component as the current residual intensity of the action. In engineering implementation, each time a new concentration sampling value and control action value are received, the control host first performs normalization, then writes them into its respective buffer, and then reads historical segments from the buffer from near to far according to preset weights and performs a weighted average to form the historical control state at the current moment. If individual sampling points show obvious abnormal jumps, they are first smoothly replaced by their immediate neighbors before entering the above calculation process. After this processing, the output historical control state... The recent concentration changes and recent control actions have been organically organized in chronological order. The next step is to directly use this state to infer the future concentration change process.

[0051] In step S102 of some embodiments, the historical control status is read. And combined with the current concentration sampling value stored in the buffer area of ​​the first step Based on this, a future predicted concentration sequence is generated. The previous step compressed recent concentration fragments and recent control action fragments into the same state chronologically; therefore, this step does not reorganize the original sampling data, but instead... This is considered a summary of the current regulatory history, from which two key quantities for future changes are extracted: the current main concentration level and the residual effects of recent control actions. In practice, this is implemented in the control unit... As a two-dimensional state vector, its first component is denoted as... The first component represents the principal concentration level formed by the average decay of historical concentration fragments; the second component is denoted as... , representing the residual intensity of motion formed by the attenuation average of historical control motion segments. Among them, This reflects the extent to which the current concentration has become apparent. This reflects how much of the subsequent effect of the preceding regulatory action will continue to be released. Based on these two quantities, a delayed release quantity is first constructed to describe the scenario where "the action has been applied but the concentration is still becoming apparent." This construction originates from the idea of ​​unreleased driving forces in first-order delayed systems: if the residual action is considered as the driving force, and the principal concentration level is considered as the already manifested proportion, then the portion that can still be released is related to both the "intensity of the residual action" and the "proportion not yet manifested." Discretizing this idea and directly applying it to the state variables of this scheme yields:

[0052] ;

[0053] in, This indicates the amount of delayed release at the current moment; Indicates the residual intensity of actions in historical regulatory states; This represents the principal concentration level in historical regulatory states. Since the historical regulatory states are constructed using a normalized weighted average, therefore... Keep it between zero and one. It remains between negative one and positive one, therefore It can be directly interpreted as the amount of action influence that will continue to be released into the future at the present moment.

[0054] In obtaining Next, the predicted concentration sequence for the future is generated. The basic form used here is derived from the discrete recursive expression of a classic first-order inertial system: the current state plus the external driving force, and the subtraction of the regression term after deviating from the equilibrium position, yields the state at the next moment. Considering the application scenario of this scheme, the "external driving force" is replaced with a future driving term composed of motion residue and delayed release, and the "equilibrium position" is replaced with the principal concentration level. The recursion starting point is set to the current concentration sample value stored in the buffer. This yields a recursive relationship applicable to isotopic gas prediction in sealed enclosures. Specifically, the calculation involves obtaining the forward... The concentration prediction value at each time step; where the concentration prediction value at the previous time step is initially the current concentration sample value; the residual intensity of the action is added to the delayed release amount, and then multiplied by the corresponding attenuation coefficient to obtain the fourth data; the concentration prediction value at the previous time step is subtracted from the main concentration level, and then multiplied by the preset regression coefficient to obtain the fifth data; the concentration prediction value at the previous time step, the fourth data, and the fifth data are added together to obtain the forward-moving concentration prediction value. The predicted concentration value at each time point. As shown in the following formula:

[0055] ;

[0056] in, Indicates the forward first The concentration prediction value at each moment is obtained by the control host through step-by-step recursion; Indicates the forward first The concentration prediction value at each time point is taken at the beginning of the recursion. ; This indicates the current concentration sample value; It represents the attenuation coefficient of the action effect as it advances with the prediction step size. Its value is determined by the concentration evolution records after a single gas injection and a single gas extraction during the system debugging phase. Indicates the residual intensity of the action; Indicates the amount of delayed release; This represents the regression coefficient, the value of which is determined during the system debugging phase based on the convergence rate when the concentration change tends to level off. Indicates the main concentration level; Indicates the prediction step number; The length of the predicted concentration sequence is preset by the system. The derivation of this recursive formula is continuous: first from... Extract the main concentration level and residual intensity of the action, and then construct the delayed release amount from these two quantities. , and then As a future driving term, it is included in the discrete first-order inertial recursion, and simultaneously used This represents the regression correction of the predicted trajectory relative to the current main concentration level. The resulting predicted concentration values ​​will not deviate infinitely along a single slope, but will first continue to change along the direction of recent action, and then gradually converge towards the vicinity of the current main concentration level.

[0057] Using a set of debugging parameters can more intuitively illustrate the calculation process. Let the historical control state be... That is, the main concentration level residual intensity of motion Meanwhile, the current concentration sampling value is During a single gas replenishment test, it was recorded that the concentration continued to rise after the control was stopped, and gradually slowed down after several cycles. Based on this, the action effect attenuation coefficient was selected. During the natural settling adjustment, a stable regression process was recorded after the concentration deviated from the main level, and the regression coefficients were selected accordingly. Predicted length is taken First, calculate the delayed release amount to obtain... Starting point of recursion When substituting into the formula to calculate the first predicted concentration, we get... Continue calculating the second concentration prediction value, and obtain... Then, calculate the third predicted concentration value forward to obtain... This set of results shows an initial significant increase, followed by a decrease in the increment and a tendency to level off, indicating that the delayed effects of recent control actions are gradually being released, and the regression term is beginning to converge. If the actual operating conditions correspond to the pumping process, then the results obtained during the commissioning phase... The trend will turn negative, and the recursive result will show an initial continued decline followed by a gradual slowdown, while maintaining consistency in the calculation process. During program implementation, only the latest data needs to be read for each adjustment cycle. and Calculate once Then, the entire predicted concentration sequence is gradually derived according to the preset prediction length. .

[0058] In some embodiments, steps S103 to S104 involve directly reading the predicted future concentration sequence. And compress this entire future trajectory into control commands for the current cycle. This approach is based on the rolling deviation evaluation concept in discrete predictive control: first, the deviations from the target value at several future moments are accumulated sequentially over time, and then the resulting comprehensive deviation is mapped to the current action. Considering that this scheme corresponds to isotope gas control in a closed chamber, the future trajectory not only has the problem of overall overestimation or underestimation, but also frequently exhibits local risks of "continuing to surge upwards in a short period of time" or "continuing to decline downwards in a short period of time." Therefore, in addition to the conventional weighted accumulation of future deviations, the extreme value risk with the largest deviation from the target value in the future predicted concentration sequence is incorporated into the current control requirements. The control host first scans the complete future predicted concentration sequence within the current cycle to find the index of the moment with the largest absolute deviation from the target value. Then, the overall deviation and the extreme risk are combined into the control demand. After this processing, the control unit does not react to a single future point, but rather reacts simultaneously to the overall deviation and the maximum deviation risk of the entire future trajectory. The previous step provided a future concentration path with delayed manifestation characteristics; this step transforms this path into the action intensity of the current cycle.

[0059] In practice, the control and solving module in the control host first scans the future predicted concentration sequence. The predicted concentration values ​​are read point by point for each future moment, and then the target concentration setpoint is read from the control interface or the experimental task configuration. The fundamental structure of future deviations originates from the discrete weighted summation form in mathematics, which is originally used to accumulate a series of discrete quantities with different weights. In this scheme, this form is applied to the deviation of future predicted concentrations relative to the target value, giving higher weight to prediction points closer to the current period and lower weight to prediction points further away, thus reflecting the solution logic of "prioritizing near-term risks." Furthermore, considering the characteristic that isotopic gases in a sealed container may continue to rise or fall due to inertia, a signed extreme value risk term is added to the weighted summation formula to incorporate the moment with the largest deviation from the target value in the future sequence into the current control demand. This yields the control demand for the current period. The calculation method is as follows: subtract the predicted concentration value from the target concentration setpoint at each time point, and then multiply by the corresponding time-series decay coefficient to obtain multiple sixth data points. Add all the sixth data points together to obtain the seventh data point. Select the largest sixth data point from all the sixth data points to obtain the eighth data point. Multiply the eighth data point by the extreme value risk coefficient and add it to the seventh data point to obtain the control demand for the current period. See the formula below:

[0060] ;

[0061] in, This indicates the control demand for the current period; Indicates the length of the future predicted concentration sequence; This represents the timing decay coefficient, which is selected during the system debugging phase based on the duration of the impact of a single control action within the enclosure. It is used to control the weight of the impact of different future moments on the current decision. Indicates the first Predicted concentration values ​​for a future time; This indicates the target concentration setpoint for the current period, which is written to the control host by the upper control interface or task script. This represents the extreme value risk coefficient, which is set during the system debugging phase based on the degree of impact of overshoot or undershoot on the experimental process. Indicates satisfaction The future moment index is obtained by the control host scanning the entire future predicted concentration sequence. The first term in the formula comes from the discrete weighted deviation integral, which accumulates the deviations of each future point relative to the target concentration setpoint in the order of closer to farther away, to obtain the overall trend of the future trajectory deviating from the target. The second term is an extreme value risk correction term added on this basis. It directly takes the prediction point with the largest deviation from the target concentration setpoint and retains its deviation sign, thus mapping the most dangerous upward or downward movement in the future to the current control requirements.

[0062] In obtaining Then, continue generating the target control instructions for the current cycle. The mapping principle used here is based on the concept of bounded proportional control in control theory. This means maintaining an approximately linear response when the deviation is small, and gradually compressing the growth rate when the deviation is large, ensuring that the control command remains sensitive while avoiding large abrupt changes. Considering the characteristic of this scheme where gas replenishment and extraction share the same control channel, the control direction is delegated to… The sign of the variable determines the control amplitude, which is then determined by a bounded fractional structure. This involves adding the preset value to the absolute value of the control demand to obtain the ninth value. Dividing the control demand by the ninth value and multiplying it by the negative of the base adjustment coefficient yields the target control command. The formula is shown below:

[0063] ;

[0064] in, This indicates the target control command for the current cycle; This represents the basic adjustment coefficient, which is calibrated during the commissioning phase based on the actual impact of a single action on concentration changes. This indicates the control demand for the current period; This represents the absolute value of the controlled demand, which is calculated directly by the program. and The same set of signed normalized control codes is used, where positive values ​​represent the direction of gas replenishment, negative values ​​represent the direction of gas extraction, and the absolute value represents the intensity of the action in the current cycle. The derivation of this formula is continuous with the previous one: the former first compresses the future trajectory into the comprehensive demand of the current cycle, and the latter then maps the comprehensive demand into bounded control commands. When When the denominator approaches zero, the denominator approaches one. and Approximately proportional change; when As the denominator increases gradually, the growth of the control command is compressed, thus keeping the current action stable.

[0065] The calculation process can be seen by combining a set of specific parameters. Let the predicted future concentration sequence be... , , , The target concentration setting value is Predicted length The time-series decay coefficient is taken as Extreme value risk coefficient is taken The future deviation weighting terms are as follows: Calculations yielded The moment when the absolute deviation of the future predicted concentration sequence from the target concentration setpoint is the largest is... Its sign bias is Extreme value risk item is Therefore, the demand is controlled to be Continue to set the basic adjustment coefficient. The target control command for the current cycle is: The denominator is 1.24175, and the fractional result is approximately 0.1947. Therefore... A negative result indicates that a directional action should be taken in the current cycle, and the intensity of this action is determined by the overall skewness of the future trajectory and the most dangerous deviation point. If the other set of predicted sequences is generally below the target value, both the weighted deviation term and the extreme value risk term turn negative, resulting in a positive result. This corresponds to the direction of air replenishment. The control unit recalculates according to this process in each control cycle. The result is then sent to the next execution module, which converts the value into valve action and pumping device action.

[0066] Furthermore, read the target control commands. This process is then transformed into the actual execution process acting on the gas flow within the box. This step is based on the execution mechanism of a discrete first-order inertial system, its original form derived from the "actuator dynamic response model" in classical control theory. This model describes how the actuator will not instantaneously reach the target control command after receiving it. Instead of directly targeting a specific value, the process gradually approaches the target value at a certain speed. In this scenario, this concept is used to drive the valves and extraction devices, making the control actions continuously change over time, thus aligning with the physical processes of gas diffusion and mixing within a confined space. Specifically, in each control cycle, the control host reads the target control command for the current cycle. And combine the actual execution control instructions from the previous cycle The actual control commands for the current cycle are generated through discrete inertial approximation. The difference between the target control instruction and the actual control instruction executed in the previous cycle is multiplied by the execution dynamic coefficient, and then added to the actual control instruction executed in the previous cycle to obtain the actual control instruction executed in the current cycle. As shown in the following formula:

[0067] ;

[0068] in, This indicates the actual execution control instructions for the current cycle; The actual execution control instructions are cached by the control host at the end of each cycle; Indicates target control commands; This represents the dynamic coefficient, determined during the system commissioning phase based on the valve response speed and the inertia of the extraction device. This formula is based on the first-order inertial response. Based on the discretization, it is obtained, where Input for the target, To execute the output, the expression in this scheme is obtained by replacing variables and introducing actual control quantities, so that the executed value gradually approaches the target control command in time. Because and All computations originate from the same control computation space and maintain a consistent scale during the recursive computation process, so the computation results can be directly used to perform mapping.

[0069] get Then, it is converted into the physical action of the specific actuator. The control host pre-stores the calibration curve of the actuator, which is obtained through experiments and is used to describe the relationship between the control signal and the actual gas flow rate. During execution, according to... The sign of the signal determines the direction of execution. A positive sign indicates the direction of gas supply, which is mapped to the valve opening degree or opening duration and sent to the solenoid valve via the drive interface. A negative sign indicates the direction of gas extraction, whose absolute value is mapped to the operating intensity or operating time of the extraction device and sent to the extraction pump drive module via the control interface. During this mapping process, the execution module converts the normalized control quantity into the corresponding physical control quantity according to the calibration curve, thereby ensuring that control signals of different amplitudes produce consistent gas flow changes in the actual system.

[0070] It should be noted that the target control command It is an instruction calculated by the control host based on the prediction results, indicating the control level that the execution end is expected to achieve in the current cycle; It refers to the actual control command issued to the valve or extraction device in the current cycle, taking into account the actuator response process. The former is the target value, and the latter is the executed value. It enables the control action to unfold smoothly according to the actual response process of the actuator, so that the valve opening or pumping intensity gradually approaches the target control level, thereby reducing local concentration disturbances, overshoot and oscillation caused by sudden action changes, and is more in line with the actual process of gas diffusion and mixing in a confined space.

[0071] The execution effect can be illustrated by a specific calculation process. Let the actual execution control instruction of the previous cycle be... The target control command for the current cycle is Execute dynamic coefficients Therefore, the actual control command executed in the current cycle is calculated as -0.12 + 0.4 × (-0.30 + 0.12) = -0.12 - 0.072 = -0.192. This result indicates that the extraction action is enhanced in this cycle, but the magnitude of the change is constrained by the execution dynamic coefficient and will not instantly reach the target value. If the same trend is maintained in the next cycle, the execution value will continue to be updated according to the same recursive relationship, so that the control action gradually approaches the target value. In this process, the execution signal changes continuously, and the gas discharge rate in the chamber changes smoothly accordingly, thus keeping the concentration change process stable.

[0072] In actual operation, the control host executes the above recursive and mapping process in each cycle, so that the control commands are gradually applied to the physical system. This step is fully utilized and transformed into continuous gas injection or discharge behavior by executing a dynamic process, so that predictive control results can be applied to the system in a manner consistent with the physical process, thereby achieving stable regulation of isotope gas concentration within the target range.

[0073] Steps S101 to S105 of this embodiment involve acquiring multiple historical concentration sampling values ​​and multiple historical control action values ​​within a historical window length, and constructing a historical control state based on these values. The historical control state includes a primary concentration level and residual action intensity. The delayed release amount is calculated based on the primary concentration level and residual action intensity. The current concentration sampling value is obtained, and multiple concentration prediction values ​​are calculated based on the current concentration sampling value, delayed release amount, primary concentration level, and residual action intensity. All concentration prediction values ​​constitute a future predicted concentration sequence. The target concentration setpoint for the current period is obtained, and the control demand for the current period is calculated based on the future predicted concentration sequence, the target concentration setpoint, a preset time-series decay coefficient, and a preset extreme value risk coefficient. The target control command for the current period is calculated based on the control demand, a preset basic adjustment coefficient, and the absolute value of the control demand. The isotope gas concentration is then adjusted according to the target control command. This improves the accuracy of isotope gas concentration control.

[0074] Please see Figure 2 This application also provides an intelligent isotope gas concentration control system, which can realize the above-mentioned intelligent isotope gas concentration control method. The system includes:

[0075] The acquisition unit 201 is used to acquire multiple historical concentration sampling values ​​and multiple historical control action values ​​of the historical window length, and construct a historical control state based on the historical concentration sampling values ​​and historical control action values; wherein, the historical control state includes the main concentration level and the residual intensity of the action;

[0076] The first calculation unit 202 is used to calculate the delayed release amount based on the main concentration level and the residual intensity of the action, obtain the current concentration sample value, and calculate multiple concentration prediction values ​​based on the current concentration sample value, the delayed release amount, the main concentration level, and the residual intensity of the action; wherein, all the concentration prediction values ​​constitute a future predicted concentration sequence;

[0077] The second calculation unit 203 is used to obtain the target concentration setpoint for the current period and calculate the control demand for the current period based on the future predicted concentration sequence, the target concentration setpoint, the preset time decay coefficient, and the preset extreme value risk coefficient.

[0078] The control unit 204 is used to calculate the target control command for the current cycle based on the control demand, the preset basic adjustment coefficient and the absolute value of the control demand, and to adjust the isotope gas concentration according to the target control command.

[0079] The specific implementation of this intelligent isotope gas concentration control system is basically the same as the specific embodiment of the intelligent isotope gas concentration control method described above, and will not be repeated here.

[0080] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for intelligent control of isotope gas concentration, characterized in that, The method includes: Multiple historical concentration sampling values ​​and multiple historical control action values ​​of a historical window length are obtained, and a historical regulation state is constructed based on the historical concentration sampling values ​​and the historical control action values; wherein, the historical regulation state includes the main concentration level and the residual intensity of the action; The delayed release amount is calculated based on the main concentration level and the residual intensity of the action, and the current concentration sample value is obtained. Multiple concentration prediction values ​​are calculated based on the current concentration sample value, the delayed release amount, the main concentration level, and the residual intensity of the action; wherein, all concentration prediction values ​​constitute the future predicted concentration sequence. Obtain the target concentration setpoint for the current period, and calculate the control demand for the current period based on the predicted future concentration sequence, the target concentration setpoint, the preset time-series decay coefficient, and the preset extreme value risk coefficient. The target control command for the current period is calculated based on the control demand, the preset basic adjustment coefficient, and the absolute value of the control demand. The isotope gas concentration is then adjusted according to the target control command.

2. The intelligent control method for isotope gas concentration according to claim 1, characterized in that, The step of adjusting the isotope gas concentration according to the target control command includes: Retrieve the actual execution control instructions from the previous cycle; The actual execution control command for the current cycle is calculated based on the actual execution control command of the previous cycle, the preset execution dynamic coefficient, and the target control command. The isotope gas concentration is adjusted according to the actual execution control command of the current cycle.

3. The intelligent control method for isotope gas concentration according to claim 2, characterized in that, The step of calculating the actual execution control command for the current cycle based on the actual execution control command of the previous cycle, the preset execution dynamic coefficient, and the target control command includes: The difference between the target control instruction and the actual execution control instruction of the previous cycle is multiplied by the execution dynamic coefficient, and then added to the actual execution control instruction of the previous cycle to obtain the actual execution control instruction of the current cycle.

4. The intelligent control method for isotope gas concentration according to claim 1, characterized in that, The step of constructing a historical control state based on the historical concentration sampling values ​​and the historical control action values ​​includes: The historical concentration sampling value and the historical control action value at each moment are concatenated and then multiplied by the corresponding memory decay coefficient to obtain multiple first data. Add all the first data together to get the second data, and add all the memory decay coefficients together to get the third data; The historical control state is obtained by calculating the ratio between the second data and the third data.

5. The intelligent control method for isotope gas concentration according to claim 1, characterized in that, The calculation of multiple concentration prediction values ​​based on the current concentration sampling value, the delayed release amount, the main concentration level, and the residual intensity of the action includes: Get forward The concentration prediction value at each time point; wherein, the concentration prediction value at the previous time point is initially the current concentration sample value; Add the residual intensity of the action to the delayed release amount, and then multiply by the corresponding attenuation coefficient to obtain the fourth data. Subtract the concentration prediction value of the previous moment from the main concentration level, and then multiply by the preset regression coefficient to obtain the fifth data. Add the predicted concentration value from the previous moment, the fourth data, and the fifth data to obtain the forward [value]. The predicted concentration value at each time point.

6. The intelligent control method for isotope gas concentration according to claim 1, characterized in that, The calculation of the control demand for the current period based on the predicted future concentration sequence, the target concentration setpoint, the preset time-series decay coefficient, and the preset extreme value risk coefficient includes: Subtract the predicted concentration value from the target concentration setpoint at each time point, and then multiply by the corresponding time-series decay coefficient to obtain multiple sixth data points. Add all the sixth data points together to obtain the seventh data point. The largest data is selected from all the sixth data to obtain the eighth data. The eighth data is multiplied by the extreme value risk coefficient and then added to the seventh data to obtain the control demand for the current period.

7. The intelligent control method for isotope gas concentration according to claim 1, characterized in that, The step of calculating the target control command for the current period based on the control demand, a preset basic adjustment coefficient, and the absolute value of the control demand includes: Add the preset number to the absolute value of the control demand to obtain the ninth data. Divide the control demand by the ninth data and multiply by the opposite of the basic adjustment coefficient to obtain the target control command.

8. An intelligent control system for isotope gas concentration, characterized in that, The system includes: An acquisition unit is used to acquire multiple historical concentration sampling values ​​and multiple historical control action values ​​for a historical window length, and to construct a historical control state based on the historical concentration sampling values ​​and the historical control action values; wherein, the historical control state includes a main concentration level and residual action intensity; The first calculation unit is used to calculate the delayed release amount based on the main concentration level and the residual intensity of the action, obtain the current concentration sample value, and calculate multiple concentration prediction values ​​based on the current concentration sample value, the delayed release amount, the main concentration level, and the residual intensity of the action; wherein, all concentration prediction values ​​constitute the future predicted concentration sequence; The second calculation unit is used to obtain the target concentration setpoint for the current period, and calculate the control demand for the current period based on the future predicted concentration sequence, the target concentration setpoint, the preset time decay coefficient, and the preset extreme value risk coefficient. The control unit is used to calculate the target control command for the current period based on the control demand, the preset basic adjustment coefficient and the absolute value of the control demand, and to adjust the isotope gas concentration according to the target control command.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the intelligent control method for isotope gas concentration as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent control method for isotope gas concentration according to any one of claims 1 to 7.

Images