A thermal coupling dynamic adaptive decoupling control method and system for a miniature atomic magnetometer

Abstract

The present application belongs to the field of quantum precision measurement, and in particular relates to a thermal coupling dynamic adaptive decoupling control method and system for a micro atomic magnetometer. The present application proposes a double-time-scale adaptive decoupling architecture: in the short time scale, a dynamic thermal coupling factor is calculated based on time delay compensation to correct the feedforward decoupling matrix in real time to suppress transient thermal disturbance; in the long time scale, a recursive least squares algorithm with a forgetting factor is used to identify and update the system parameters online and correct the decoupling matrix reference value. The fast and slow loops run in parallel with different time granularities and jointly act on the same decoupling matrix. The present application realizes a ±0.01℃ steady-state temperature control precision, shortens the start-up lock-in time to 1 / 10 of the traditional method, and has a 72h long-term error of better than 0.021℃, significantly improving the measurement precision and long-term stability of the atomic magnetometer.

Description

Technical Field

[0001] This invention belongs to the field of quantum precision measurement and sensor temperature control technology, specifically relating to a thermal coupling dynamic adaptive decoupling control method and system for a micro atomic magnetometer. It is mainly used to solve the problems of insufficient temperature control accuracy, long start-up lock-up time and poor long-term operation stability caused by strong thermal coupling between the atomic gas chamber and the laser. Background Technology

[0002] Miniature atomic magnetometers (such as SERF magnetometers or chip-scale atomic magnetometers CSAM) rely on the spin effect of alkali metal atoms to achieve ultra-high precision magnetic field measurement. As systems develop towards miniaturization and integration, the internal space of the equipment is highly compact, the operating temperature difference between the atomic gas chamber and the laser can reach 100°C, and the spatial distance between the two is only on the millimeter level, resulting in extremely strong thermal interaction.

[0003] In existing technologies, the conventional temperature control range for atomic gas chambers is 140℃~180℃, and the conventional temperature control range for lasers is 50℃~90℃. The close-range coupling of high and low temperature modules presents several technical drawbacks, as follows:

[0004] 1) The thermal coupling effect between the gas chamber and the laser is significant. The high temperature of the gas chamber will continuously generate thermal disturbance to the laser, causing frequent temperature fluctuations in the laser and making it difficult to guarantee the temperature control accuracy.

[0005] 2) The internal heat conduction of the equipment has asymmetric time delay characteristics, which is insufficient to compensate for by traditional control strategies. Overshoot and oscillation are prone to occur during the temperature regulation process, resulting in poor temperature control stability.

[0006] 3) During long-term operation of the equipment, the aging of components and the decay of vacuum will cause the thermodynamic parameters to drift, resulting in a mismatch between the preset temperature control model and the actual working conditions, and the overall stability of the system will continue to decline.

[0007] 4) During the system startup phase, the laser is affected by the thermal shock of the gas chamber, which can cause severe temperature overshoot. The steady-state lock-in time generally exceeds 20 minutes, resulting in slow equipment startup response.

[0008] 5) Under steady-state operating conditions, even small temperature fluctuations can directly cause the laser's operating wavelength to drift, significantly reducing the long-term magnetic field measurement accuracy and data reliability of the atomic magnetometer.

[0009] like Figure 1 As shown, since the distance between the atomic gas chamber and the laser module is only on the order of millimeters, the thermal radiation generated by the high-temperature gas chamber, the air heat conduction, and the PCB heat conduction path together form a strong thermal coupling effect.

[0010] Currently, there are some related technical solutions in the field of temperature control for atomic magnetometers. For example, Chinese patent CN115857588A discloses a temperature control system, method, device, and medium for the gas chamber of a SERF atomic magnetometer, which uses a photodiode light intensity signal as feedback and a PID control algorithm to drive a non-magnetic heating element; Chinese patent CN116594447A discloses a temperature stabilization control system and method for the atomic gas chamber of a SERF magnetometer, which uses the ratio of a second-order demodulated signal to a DC signal as the temperature signal; Chinese patent CN120603080A discloses a temperature control circuit and an atomic magnetometer, including a PID control module; and Chinese patent CN121440360B discloses a closed-loop temperature compensation circuit and temperature compensation method for the laser of an atomic magnetometer. However, the above-mentioned existing technologies all adopt single-loop independent PID control or passive heat insulation strategies, which fail to effectively solve the problem of bidirectional thermal coupling between the gas chamber and the laser.

[0011] In the field of industrial temperature control, although some multi-channel decoupled temperature control technologies already exist, such as the multi-channel decoupled temperature control method disclosed in Chinese patent CN104656699A, which achieves multi-channel decoupling by constructing a transfer function matrix and designing a decoupling network; the PID temperature control system and method for semiconductor etching machines disclosed in Chinese patent CN120909373A, which achieves decoupled control of multi-region temperature control by constructing a coupling relationship matrix and its inverse matrix, and adopts the correlation between the response residual vector and the control vector, dynamically optimizes the coupling matrix, and sets a rapid response mechanism for transient temperature changes; and the multivariable decoupling method for the temperature control system of testing equipment disclosed in Chinese patent CN121143545A, which realizes online adaptive identification and compensation of coupling characteristics and can automatically adapt to the time-varying coupling characteristics caused by equipment aging.

[0012] However, the aforementioned MIMO decoupling solutions and FF-RLS applications in the field of industrial temperature control all have the following unresolved technical gaps:

[0013] First, the controlled object is a highly symmetrical industrial multi-temperature zone, which does not have the special physical constraints of an extreme temperature difference of nearly 100°C and a spatial distance of millimeters between the gas chamber (140°C~180°C) and the laser (50°C~90°C) in the atomic magnetometer.

[0014] Secondly, the asymmetric bidirectional thermal coupling characteristics were not characterized. The coupling between channels in industrial multi-temperature zones usually has good symmetry. However, the coupling strength between the gas cell and the laser in the atomic magnetometer is much greater than that between the laser and the gas cell, and the time delay difference is more significant (the pure lag between the gas cell and the laser is about 35 seconds, while the pure lag between the main laser channel and the gas cell is only about 3 seconds).

[0015] Third, although existing solutions mention dynamic optimization of coupling matrix (CN120909373A) or online adaptive identification (CN121143545A), neither of them performs dedicated time shift compensation for asymmetric delay, nor do they establish a cooperative control architecture with fast and slow dual time scales to clarify the hierarchical parallelism of different time granularities. The transient fast response and adaptation of existing solutions are usually functional descriptions under the same time scale, rather than architecturally hierarchically layering the two as two loops that run independently and in parallel with clear different time granularities.

[0016] Fourth, none of the existing solutions disclose the specific calculation method of the dynamic thermal coupling factor based on time delay compensation, especially the specific algorithm that combines the time shift compensation factor with the sliding window integral smoothing.

[0017] In summary, there is an urgent need in this field for a temperature control method and system for atomic magnetometers that can simultaneously solve the three core problems of strong thermal coupling crosstalk, asymmetric time delay compensation lag, and long-period parameter drift. Summary of the Invention

[0018] This invention aims to solve the industry problem of strong thermal coupling crosstalk between the atomic gas chamber and the laser in the extreme conditions of compact space and large temperature difference in miniature atomic magnetometers, which makes it impossible to achieve high-precision temperature control with traditional single-loop control and passive heat insulation strategies. At the same time, it effectively optimizes the system startup response speed and significantly improves the stability and measurement accuracy of the equipment during long-term continuous operation.

[0019] This invention proposes an integrated solution for thermally coupled modeling, feedforward decoupling, time delay compensation, and dual-timescale online adaptation. Its core lies in constructing a dual-timescale (variable timescale) adaptive architecture.

[0020] For ease of description, the short-timescale dynamic correction mechanism is referred to as the fast loop, and the long-timescale online identification mechanism is referred to as the slow loop.

[0021] The fundamental difference between this architecture and existing technologies lies in the fact that existing technologies (such as CN120909373A) typically integrate transient temperature change rapid response mechanisms and self-learning and model adaptation capabilities as complementary functions operating within the same control framework and at the same time granularity. They do not explicitly distinguish these two aspects at the architectural level as two independent loops operating in parallel with different time granularities. This invention creatively recognizes that transient thermal shocks (second-level) and long-period parameter drifts (hour-level to day-level) occur on completely different time scales and should be handled separately with different time granularities.

[0022] Under short time scales (fast loop), the system operates with the sampling period (second level) as the time granularity. By introducing a time delay compensation factor to calculate the dynamic thermal coupling factor, the feedforward decoupling matrix is ​​rapidly fine-tuned to dynamically offset the thermal shock caused by transient large-scale temperature changes or sudden disturbances.

[0023] On long time scales (slow loops), the algorithm operates at the minute-level time granularity and introduces a recursive least squares (FF-RLS) algorithm with a forgetting factor to iteratively update the slowly changing thermodynamic parameters of the system in real time.

[0024] The fast and slow loops operate in parallel and cooperate with each other at clearly defined time granularities, working together on the same feedforward decoupling matrix. The fast loop corrects the instantaneous value of the decoupling parameter, while the slow loop corrects the reference value of the decoupling parameter, forming a collaborative mechanism for instantaneous fine-tuning and long-term calibration. This is an architectural innovation that has not been disclosed in existing technologies.

[0025] The technical solution of the present invention is as follows:

[0026] In a first aspect, a thermally coupled dynamic adaptive decoupling control method for a micro atomic magnetometer includes:

[0027] A thermal coupling transfer function model between the atomic gas chamber and the laser is established. The thermal coupling transfer function model includes the thermal temperature control main channel transfer function of the atomic gas chamber, the thermal temperature control main channel transfer function of the laser, the thermal disturbance coupling channel transfer function of the atomic gas chamber to the laser, and the thermal disturbance coupling channel transfer function of the laser to the atomic gas chamber.

[0028] Based on the thermal coupling transfer function model, a feedforward decoupling matrix is ​​designed to convert the virtual control quantity into the actual heater control output quantity.

[0029] On a short timescale, with the system sampling period Δt as the time granularity, the dynamic thermal coupling factor is calculated based on the pure time delay compensation of the thermal disturbance channel of the laser by the atomic gas cell, and the feedforward decoupling matrix is ​​corrected in real time to suppress transient thermal disturbances; wherein, the value range of the sampling period Δt is 0.1~10 seconds;

[0030] On a long time scale, with a time granularity of 30 to 300 seconds (minutes), a recursive least squares algorithm with a forgetting factor is used to update the system's thermodynamic parameters online and simultaneously correct the baseline value of the feedforward decoupling matrix; wherein, the value range of the minute-level time granularity is 30 to 300 seconds.

[0031] The real-time correction of the dynamic thermal coupling factor at the short time scale and the online update of the recursive least squares algorithm at the long time scale run in parallel with different time granularities, and both act on the same feedforward decoupling matrix. The time granularity of the long time scale is at least 60 times that of the time granularity of the short time scale.

[0032] A further improvement, namely the calculation of the dynamic thermal coupling factor and real-time correction of the feedforward decoupling matrix based on the pure time delay compensation of the laser thermal perturbation channel under the atomic gas cell at a short time scale, includes:

[0033] Calculate real-time thermal coupling strength index :

[0034] ;

[0035] in, This represents the temperature change of the laser at time k. This represents the temperature change of the atomic gas chamber after thermal coupling delay compensation. Thermal coupling channel pure time delay The corresponding number of discrete sampling steps; The system sampling period; Let K be the real-time temperature of the laser at time k. for Real-time temperature of the atomic gas chamber;

[0036] The smoothed real-time thermal coupling factor is calculated using a sliding window integration method with a dead zone. :

[0037] ;

[0038] in, The length of the sliding window; A regularization factor is used to prevent the denominator from approaching zero;

[0039] The decoupling compensation parameters in the feedforward decoupling matrix are corrected in real time based on the smoothed real-time thermal coupling factor.

[0040] ;

[0041] in: The nominal decoupling parameters identified during system initialization; This is the reference thermal coupling factor under rated operating conditions; To adaptively adjust the gain coefficient; These are the decoupling compensation parameters updated in real time.

[0042] A further improvement is made to the thermal coupling transfer function model, where the pure time τ of the atomic gas cell on the thermal perturbation coupling channel of the laser is... 21 The pure time delay τ of the laser master control channel is greater than 22 Furthermore, the steady-state gain K of the atomic gas cell on the thermal perturbation coupling channel of the laser is... 21The steady-state gain K of the laser on the thermal disturbance coupling channel of the gas chamber is greater than that of the gas chamber. 12 .

[0043] A further improvement is made to the feedforward decoupling matrix, which is:

[0044] ;

[0045] The formula for converting virtual control quantity to actual heater control output quantity is:

[0046]

[0047] in, This is a virtual control quantity for the atomic gas chamber. This is a virtual control variable for the laser; This is the actual control output of the atomic gas chamber. This refers to the actual control output of the laser heater;

[0048] G 11 (s) is the transfer function of the main channel for room temperature control of atomic gas, G 22 (s) is the transfer function of the main channel for laser temperature control, G 12 (s) is the transfer function of the laser to the thermal perturbation coupling channel of the atomic gas cell, G 21 (s) is the transfer function of the thermal perturbation coupling channel of the atomic gas cell to the laser.

[0049] A further improvement involves discretizing the feedforward decoupling matrix using a bilinear transformation and combining it with a historical control quantity buffer FIFO queue (First In First Out, a linear data structure) determined by the time delay difference to achieve advance hedging control in the discrete domain; the time delay difference includes the pure time τ of the gas chamber's response to the laser's thermal disturbance channel. 21 With the pure time delay τ of the laser master control channel 22 difference.

[0050] A further improvement is made to the recursive least squares algorithm with a forgetting factor, using a fixed time interval T. RLS Once started, T RLS The value range is 30~300 seconds. The regression vector contains the temperature measurement values ​​and control output values ​​of multiple sampling times before the current time. The recursive least squares algorithm with forgetting factor updates the system parameter vector containing the steady-state gain and time constant of each channel online.

[0051] Secondly, a thermally coupled dynamic adaptive decoupling control system for a micro atomic magnetometer includes:

[0052] A temperature control loop for atomic gas chambers is used to collect the temperature of the atomic gas chambers and generate virtual control quantities for the atomic gas chambers based on the deviation between the setpoint and the measured value.

[0053] The laser temperature control loop is used to collect the laser temperature and generate a virtual control quantity for the laser based on the deviation between the temperature setpoint and the measured value.

[0054] The feedforward decoupling module is used to design a feedforward decoupling matrix based on the thermal coupling transfer function model between the atomic gas chamber and the laser, and to convert the virtual control quantities of the atomic gas chamber and the laser into the actual control output quantities of the atomic gas chamber heater and the laser heater.

[0055] The time delay compensation module is used to determine the depth of the historical control quantity buffer FIFO queue based on the asymmetric time delay difference between the laser thermal disturbance channel and the laser main control channel in the atomic gas cell, so as to realize the advance hedging control in the discrete domain.

[0056] The short-timescale dynamic correction module is used to calculate the dynamic thermal coupling factor and correct the feedforward decoupling matrix in real time based on the pure time delay compensation of the thermal disturbance channel of the laser by the atomic gas cell with the system sampling period Δt as the time granularity. The value range of the sampling period Δt is 0.1~10 seconds.

[0057] The long-term online identification and update module is used to update the system thermodynamic parameters online with a minute-level time granularity and a recursive least squares algorithm with a forgetting factor, and simultaneously correct the reference value of the feedforward decoupling matrix. The minute-level time granularity ranges from 30 to 300 seconds.

[0058] The heating drive module is used to drive the atomic gas chamber heater and the laser heater according to the actual control output.

[0059] A further improvement is made to the atomic gas room temperature control loop, which includes:

[0060] The first temperature sensor is located in the atomic gas chamber and is used to collect the real-time temperature of the atomic gas chamber.

[0061] The first virtual controller is used to generate virtual control quantities for the atomic gas chamber based on the deviation between the setpoint temperature of the atomic gas chamber and the real-time temperature of the atomic gas chamber.

[0062] A further improvement is made to the laser temperature control circuit, which includes:

[0063] The second temperature sensor is located at the laser and is used to collect the real-time temperature of the laser.

[0064] The second virtual controller is used to generate a virtual control quantity for the laser based on the deviation between the laser temperature setpoint and the real-time temperature of the laser.

[0065] The short-timescale dynamic correction module and the long-timescale online identification and update module operate in parallel with different time granularities to jointly correct the same feedforward decoupling matrix, and the time granularity of the long-timescale is at least 60 times that of the short-timescale.

[0066] In a further improvement, the delay compensation module determines the buffer depth of the FIFO queue based on the delay difference. The FIFO queue is used to store historical control quantities. When performing discrete domain decoupling operations, the feedforward decoupling module reads historical control quantities from the FIFO queue to achieve advance hedging control.

[0067] The essential difference between this invention and existing technologies (such as CN120909373A and CN121143545A) lies in the fact that the transient temperature change rapid response mechanism and self-learning and model adaptation capabilities of existing technologies are usually complementary functions operating at the same time granularity under the same control framework, without clearly distinguishing them as two independent loops operating in parallel at different time granularities at the architectural level. This invention creatively constructs a hierarchical collaborative architecture with a fast loop (sampling period level) and a slow loop (minute level) operating in parallel at clearly different time granularities, both acting on the same feedforward decoupling matrix. The fast loop corrects the instantaneous values ​​of the decoupling parameters, while the slow loop corrects the baseline values ​​of the decoupling parameters, forming a collaborative mechanism of instantaneous fine-tuning + long-term calibration. This explicit hierarchical understanding of time scale and its specific implementation are not disclosed in existing technologies.

[0068] It should be noted that although CN120909373A mentions dynamically optimizing the coupling matrix, it does not disclose a method based on asymmetric time delay compensation. The specific method for calculating the coupling strength is not disclosed, nor are the specific implementations of the sliding window integral smoothing and the regularization factor to prevent the denominator from approaching zero. This invention is based on the pure time lag τ between the atomic magnetometer gas cell and the laser channel. 21 Much larger than the laser main channel τ 22 This asymmetric time delay characteristic introduces a time shift compensation factor, and the smoothed real-time thermal coupling factor is calculated through sliding window integration. This algorithm achieves accurate real-time sensing of asymmetric thermal coupling strength and rapid online correction of the decoupling matrix. The specific implementation of this algorithm is not publicly disclosed under the existing technology (CN120909373A) which uses the concept of dynamically optimizing the coupling matrix.

[0069] It should be noted that the atomic gas temperature control loop and the laser temperature control loop in this invention differ from traditional single-loop independent temperature control schemes. In traditional schemes, the two loops operate completely independently and do not communicate with each other; temperature fluctuations in the gas chamber are transmitted through physical thermal coupling paths (such as...). Figure 1The three paths shown passively affect the laser temperature, while the laser temperature control loop only responds passively after a temperature deviation occurs. In this invention, the virtual control quantities V1 and V2 generated by the gas chamber temperature control loop and the laser temperature control loop do not directly drive the heater. Instead, they are first sent to the feedforward decoupling module, converted into actual control output quantities U1 and U2 by the decoupling matrix D(s), and then drive the heater. This conversion process ensures that the heating demand V1 of the gas chamber has already generated a reverse compensation action on the laser heater through the decoupling matrix before affecting the laser, achieving active decoupling rather than a passive response.

[0070] It should be noted that the time delay compensation module in this invention is fundamentally different from the pure time delay processing in conventional digital control. Conventional methods typically only add a fixed delay element to the controller to match the system time delay, without considering the asymmetric time delay differences in multivariable systems. The time delay compensation module in this invention is specifically designed for the pure time delay τ between the atomic gas cell and the laser channel in an atomic magnetometer. 21 =35s is much greater than the laser main channel τ 22 The asymmetric property of =3s is based on the difference (τ) 21 -τ 22 Quantitatively determine the buffer depth of the FIFO queue so that the feedforward decoupling module can read (τ) from the FIFO queue when performing matrix multiplication. 21 -τ 22 The historical control quantity before time τ, thus before the thermal disturbance actually reaches the laser (τ). 21 -τ 22 The reverse compensation action is completed in seconds. This asymmetric time delay difference-driven advance hedging mechanism is not disclosed in existing technology.

[0071] Compared with the prior art, the beneficial effects of this invention are as follows:

[0072] 1. This invention achieves ultra-high precision steady-state temperature control by significantly suppressing bidirectional thermal crosstalk interference through feedforward decoupling and real-time correction of the dynamic thermal coupling factor based on time delay compensation. Even under extreme conditions where the gas chamber experiences a large temperature fluctuation of ±5℃, the laser temperature fluctuation can be controlled within 0.03℃, and the system's short-term steady-state temperature control accuracy can reach ±0.01℃ (with a long-term cumulative error of better than 0.021℃ over 72 hours).

[0073] 2. This invention can significantly shorten the system startup lock-up time; relying on the real-time correction of the dynamic thermal coupling factor of the fast loop with a short time scale, it can detect and compensate for temperature disturbances caused by thermal shock in the gas chamber in advance, significantly suppressing the laser temperature overshoot phenomenon. Experimental tests show that this invention can significantly shorten the system steady-state lock-up time from more than 20 minutes in traditional PID control to less than 2 minutes.

[0074] 3. This invention utilizes online adaptive iteration of FF-RLS in a long-term slow-loop system, enabling the system to detect thermodynamic parameter drift in real time and dynamically update the decoupling matrix reference value. In long-term operational experiments, even with thermal coupling gain changes exceeding 15%, the system maintains decoupling control, with laser temperature fluctuations increasing by no more than 0.005℃.

[0075] 4. This invention exhibits a synergistic effect across two time scales: the fast loop (sampling period level) and the slow loop (minute level) operate in parallel with clearly defined time granularities, respectively handling transient disturbances and long-term drift, while both jointly correct the same decoupling matrix. Comparative experiments show that with only the fast loop, the error deteriorates to 0.08℃ after 72 hours; with only the slow loop, the startup stabilization time is as long as 18 minutes; while this invention, through fast-slow synergy, simultaneously achieves 1.5-minute stabilization, 0.028℃ disturbance suppression, and 0.021℃ / 72-hour long-term accuracy, fully demonstrating the unexpected technical effects brought about by the architectural innovation of separating and synergizing the time scales of the fast and slow loops. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the thermal coupling between the atomic gas chamber and the laser.

[0077] Figure 2 This is a schematic diagram of the thermal coupling relationship between two inputs and two outputs.

[0078] Figure 3 A schematic diagram of the system step response curve and FOPDT parameter extraction;

[0079] Figure 4 This is the overall control block diagram of the thermally coupled dynamic adaptive decoupling control system;

[0080] Figure 5 A comparison chart of disturbance suppression effects (comparison between this invention and traditional PID);

[0081] Figure 6 The diagram shows the convergence process of online parameter identification for FF-RLS. Detailed Implementation

[0082] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0083] In the description of this invention, it should be noted that, unless otherwise stated, "a plurality of" means two or more; the terms "upper," "lower," "left," "right," "inner," "outer," "front end," "rear end," "head," "tail," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0084] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0085] Example 1

[0086] 1. Modeling of MIMO thermal coupling

[0087] The atomic gas chamber and the laser are integrated into a dual-input dual-output thermally coupled control system. The system inputs are the atomic gas chamber heater control input and the laser heater control input, and the outputs are the real-time temperature of the atomic gas chamber and the real-time temperature of the laser.

[0088] The system transfer function matrix is ​​in the form of a 2x2 square matrix, as follows:

[0089] ;

[0090] In the formula, The transfer function for the room temperature controlled main channel of atomic gas; For the main channel transfer function of the laser temperature control; represents the thermal disturbance coupling channel transfer function of the laser to the atomic gas cell; This is the transfer function of the thermal disturbance coupling channel of the atomic gas cell to the laser.

[0091] Each temperature control channel is accurately modeled using a first-order inertial plus pure time delay (FOPDT) model. The formula for the single-channel model is as follows:

[0092] ;

[0093] In the formula: The steady-state gain of the system reflects the steady-state temperature rise of the temperature control channel. The system time constant reflects the temperature control response speed; is the pure time delay, reflecting the heat flow propagation delay characteristics of the system; s is the Laplace operator.

[0094] like Figure 2 As shown, this invention integrates the atomic gas chamber and the laser into a dual-input, dual-output thermally coupled control system. The input u1 is the control input for the atomic gas chamber heater (gas chamber heating power), and u2 is the control input for the laser heater (laser TEC power); the output y1 is the real-time temperature of the atomic gas chamber, and y2 is the real-time temperature of the laser; G 11 (s) and G 22 (s) is the transfer function of the master control channel, G 12 (s) and G 21 (s) represents the transfer function of the thermally coupled perturbation channel. Each channel uses the transfer function G. ij (s) indicates that the arrow direction indicates the signal flow. Wherein, G ij (s) represents the transfer characteristic of the j-th input to the i-th output. j is the column index, representing the input signal number; i is the row index, representing the output signal number.

[0095] This model can fully characterize the asymmetric thermal coupling and differential time delay characteristics between the atomic gas chamber and the laser.

[0096] 2. System Parameter Identification

[0097] In the initialization phase, this invention applies a fixed step power excitation signal to the atomic gas chamber heater and the laser heater respectively, simultaneously and in real-time collects the temperature dynamic response data of the two devices, and uses the least squares method to fit the FOPDT model to accurately identify the initial values ​​of the model parameters for each channel. , , This provides a precise initial model foundation for subsequent decoupling control.

[0098] like Figure 3 As shown, the system response curve was obtained through a step excitation experiment, and relevant static parameters were extracted. During the system initialization phase, a step power excitation signal was applied to the heater (…). Figure 3 (See the sub-figure below) Simultaneously collect dynamic temperature response data ( Figure 3 (See the sub-figure above). The response curve exhibits typical FOPDT characteristics: a pure time delay stage (0~τ), an inertial stage (τ~τ+T), and a steady-state stage. Three key parameters can be extracted from this curve: steady-state gain K, time constant T, and pure time delay τ.

[0099] 3. Short timescale (fast loop): Calculation and transient correction of dynamic thermal coupling factor

[0100] To cope with transient strong interference caused by system startup or large external temperature disturbances, the system operates a dynamic thermal coupling factor correction mechanism with the system sampling period Δt as the time granularity under a short time scale (fast loop). The value of the sampling period Δt ranges from 0.1 to 10 seconds.

[0101] Considering the significant asymmetric thermal lag time between the atomic gas cell and the laser, a time-shift compensation factor is introduced, and the real-time thermal coupling strength index is... The calculation formula is as follows:

[0102] ;

[0103] in: This represents the temperature change of the laser at time k. This represents the temperature change of the atomic gas chamber after thermal coupling delay compensation. Thermal coupling channel pure time delay The corresponding number of discrete sampling steps; This is the system sampling period. Let K be the real-time temperature of the laser at time k. for The real-time temperature of the atomic gas chamber.

[0104] It should be noted that while existing technologies (such as CN120909373A) mention dynamically optimizing the coupling matrix, they do not disclose methods based on asymmetric delay compensation. The specific method for calculating the coupling strength. In particular, the pure time delay τ from the atomic gas cell to the laser channel in the atomic magnetometer. 21 τ is much larger than that of the main laser channel. 22 This asymmetric time delay characteristic is a unique physical constraint in the temperature control scenario of atomic magnetometers, and existing MIMO decoupling solutions in the field of industrial temperature control have not carried out special time shift compensation processing for this.

[0105] To reduce measurement noise and prevent numerical dispersion when the denominator approaches zero, a sliding window integration method with a dead zone is used to calculate the smoothed real-time thermal coupling factor. :

[0106] ;

[0107] in: The length of the sliding window; A regularization factor to prevent the denominator from approaching zero.

[0108] The above calculations can reflect the influence of the temperature change of the atomic gas chamber on the temperature change of the laser under the current operating conditions, thereby characterizing the dynamic change of the thermal coupling strength of the system.

[0109] Subsequently, the feedforward decoupling matrix is ​​corrected online based on the smoothed real-time thermal coupling factor:

[0110] ;

[0111] in: The nominal decoupling parameters identified during system initialization; This is the reference thermal coupling factor under rated operating conditions; To adaptively adjust the gain coefficient; These are the decoupling compensation parameters updated in real time.

[0112] When the thermal coupling between the atomic gas cell and the laser is enhanced, the thermal coupling factor increases, and the system automatically increases the feedforward compensation intensity; when the thermal coupling is weakened, the compensation amount is reduced accordingly, thereby ensuring that the decoupling control effect is always maintained in the optimal state.

[0113] Through the above-mentioned dynamic thermal coupling factor estimation and decoupling matrix adaptive correction mechanism, the present invention can compensate for the thermal coupling characteristic drift caused by device aging, packaging stress change, vacuum degree decay and ambient temperature change in real time without frequently re-identifying the complete thermal model, which significantly improves the long-term operating stability and temperature control accuracy of the system.

[0114] 4. Feedforward decoupling and time delay compensation

[0115] Based on the established MIMO thermal coupling model, a cross-feedforward decoupling matrix is ​​designed to counteract bidirectional thermal coupling interference. The form of the feedforward decoupling matrix is ​​as follows:

[0116] ;

[0117] By combining virtual control input with a feedforward decoupling matrix, the conversion from virtual control command to actual heater output command is achieved. The actual output conversion formula is as follows:

[0118]

[0119] In the formula: This is a virtual control quantity for the atomic gas chamber. This is a virtual control variable for the laser; This is the actual control output of the atomic gas chamber. This represents the actual control output of the laser heater.

[0120] Because the system satisfies and The decoupling matrix is ​​physically fully realizable. In the digital control domain, the decoupling matrix is ​​discretized through a bilinear transformation and combined with time delay differences (such as...). A FIFO queue of historical control values ​​is used to achieve advance hedging control in the discrete domain, completely eliminating overshoot and oscillation caused by adjustment lag. The time delay difference includes the pure lag time of the gas chamber on the laser thermal disturbance channel. Pure time delay of the laser main channel difference.

[0121] 5. Long-term scale (slow loop): FF-RLS online adaptive correction

[0122] To address the long-period, slowly varying parameter drift caused by component aging and cavity vacuum decay during long-term equipment operation, a recursive least squares (FF-RLS) algorithm with a forgetting factor is introduced at the minute-level granularity on a long timescale (slow loop). After dynamically updating the continuous-time model parameters, the algorithm synchronously corrects the baseline value of the decoupling matrix. The time granularity at the minute level ranges from 30 to 300 seconds, and the time granularity at the long time scale is at least 60 times that at the short time scale.

[0123] The core recursive formula of the algorithm is as follows:

[0124] ;

[0125] ;

[0126] ;

[0127] In the formula: This is a system parameter vector composed of the slowly varying thermal resistance, time constant, and steady-state gain of each channel; It is a regression vector composed of historical temperature data and historical control output signals; This is the real-time temperature measurement value at the current moment; It is the covariance matrix; is the forgetting factor, with a value range of 0.98 to 0.995; K(k) is the gain matrix. Figure 6 The convergence process of the thermally coupled parameters during online identification is presented, showing that the estimated parameter values ​​gradually approach the true values ​​and remain stable.

[0128] like Figure 6 As shown, the horizontal axis represents time (hours), and the vertical axis represents the thermal coupling channel G. 21 The steady-state gain K of (s) 21The estimated value (normalized per unit value) is as follows: 0-5 hours is the identification and convergence stage (fluctuation of about ±5%), 5-20 hours gradually converges and tends to stabilize, and 20-72 hours stabilizes near the true value (fluctuation < ±1%). The dashed line represents the slow change trend of the true value caused by device aging. The FF-RLS estimate can effectively track this trend.

[0129] Preferably, the recursive least squares algorithm uses a fixed time interval T. RLS Once started, T RLS The value range is 30~300 seconds. Preferably, the regression vector... It includes temperature measurements and control output values ​​from multiple sampling times prior to the current time. Preferably, the algorithm updates the steady-state gain K of each channel online. ij and time constant T ij System parameter vector .

[0130] It should be noted that the FF-RLS algorithm itself is a known method in the field of system identification. The contribution of this invention does not lie in the FF-RLS algorithm itself, but in:

[0131] First, FF-RLS is applied for the first time to the parameter drift compensation unique to atomic magnetometers. The drift mechanism and rate of change of the slowly changing thermodynamic parameters caused by device aging and vacuum chamber decay are completely different from other temperature control scenarios.

[0132] Second, FF-RLS does not operate independently in this invention, but is used as a slow loop in a dual-timescale architecture with a time granularity of minutes. It is clearly distinguished from, runs in parallel with, and works collaboratively with the fast loop (sampling period level). This collaborative use method with explicit timescale layering has not been disclosed in the prior art.

[0133] This invention constructs a thermally coupled dynamic adaptive decoupling control system for a micro atomic magnetometer, including an atomic gas temperature control loop, a laser temperature control loop, a feedforward decoupling module, a time delay compensation module, a long-timescale online identification module, a short-timescale dynamic correction module, and a heating drive module.

[0134] (1) Atomic gas room temperature control loop

[0135] The atomic gas room temperature control loop is used to acquire the temperature of the atomic gas chamber and generate a virtual control quantity for the atomic gas chamber based on the deviation between the setpoint and the measured value. Specifically, the atomic gas room temperature control loop includes:

[0136] The first temperature sensor is located in the atomic gas chamber and is used to collect the real-time temperature y1(K) of the atomic gas chamber.

[0137] The first virtual controller (such as a PID controller) is used to generate the virtual control quantity V1(k) of the atomic gas chamber based on the deviation e1(k) = r1 - y1(k) between the setpoint r1 of the atomic gas chamber temperature and the real-time temperature y1(k) of the atomic gas chamber. The parameters of the first virtual controller can be designed independently without considering the thermal coupling effect between the atomic gas chamber and the laser, because the coupling effect will be handled by the subsequent feedforward decoupling module.

[0138] (2) Laser temperature control circuit

[0139] The laser temperature control circuit is used to acquire the laser temperature and generate a virtual control quantity for the laser based on the deviation between the setpoint and the measured temperature. Specifically, the laser temperature control circuit includes:

[0140] The second temperature sensor is located at the laser and is used to collect the real-time temperature y2(k) of the laser.

[0141] The second virtual controller (such as a PID controller) is used to generate a virtual control quantity V2(k) for the laser based on the deviation e2(k) = r2 - y2(k) between the laser temperature setpoint r2 and the real-time laser temperature y2(k). Similarly, the parameters of the second virtual controller can be designed independently.

[0142] (3) Feedforward decoupling module

[0143] The feedforward decoupling module is used to design the feedforward decoupling matrix D(s) based on the thermal coupling transfer function model between the atomic gas chamber and the laser. This matrix converts the virtual control quantities V1(k) of the atomic gas chamber and V2(k) of the laser into the actual control output quantities U1(k) and U2(k) of the atomic gas chamber heater and the laser heater, respectively. The conversion formula and the form of the feedforward decoupling matrix D(s) are described above and will not be repeated here.

[0144] (4) Delay compensation module

[0145] The time delay compensation module is used to determine the depth of the historical control quantity buffer FIFO queue based on the asymmetric time delay difference between the gas chamber's thermal disturbance channel and the laser's main control channel, thereby achieving advance hedging control in the discrete domain. Specifically, the time delay compensation module is based on the time delay difference τ. 21 -τ 22 The buffer depth of the FIFO queue is determined, and the FIFO queue is used to store historical control values. When performing discrete domain decoupling operations, the feedforward decoupling module reads historical control values ​​from the FIFO queue to achieve advance counter-compensation control by applying reverse compensation actions before the disturbance actually reaches the controlled object.

[0146] Due to the pure time τ of the thermal disturbance channel of the laser in the gas cell of the atomic magnetometer... 21=35s is much greater than the pure time delay τ of the laser master control channel. 22 =3s, the delay compensation module calculates the difference (τ) 21 -τ 22 Set the FIFO queue depth to ensure that the decoupling compensation signal can be delivered in advance (τ). 21 -τ 22 Time is applied to the laser heater to complete the hedging before the thermal disturbance actually affects the laser temperature.

[0147] (5) Short-timescale dynamic correction module (fast loop)

[0148] The short-timescale dynamic correction module is used to calculate the dynamic thermal coupling factor based on the pure time delay compensation of the gas chamber on the thermal disturbance channel of the laser, with the system sampling period Δt as the time granularity, and to correct the feedforward decoupling matrix in real time. The value range of the sampling period Δt is 0.1~10 seconds.

[0149] (6) Long-term online identification module (slow loop)

[0150] The long-time scale online identification module is used to update the system thermodynamic parameters online with a recursive least squares algorithm with a forgetting factor at a time granularity of minutes, and simultaneously correct the reference value of the feedforward decoupling matrix. The time granularity of minutes ranges from 30 to 300 seconds.

[0151] (7) Heating drive module

[0152] The heating drive module is used to drive the gas chamber heater and the laser heater according to the actual control outputs U1(k) and U2(k). Preferably, the gas chamber heater is a non-magnetic heating element disposed on the outer wall of the atomic gas chamber; the laser heater is a TEC cooling element disposed on the laser module.

[0153] (8) Collaborative relationships between modules

[0154] The short-timescale dynamic correction module (fast loop) and the long-timescale online identification module (slow loop) operate in parallel with different time granularities, jointly correcting the same feedforward decoupling matrix, and the time granularity of the long-timescale is at least 60 times that of the short-timescale. Specifically:

[0155] The fast loop uses the sampling period (on the order of seconds) as the time granularity and calculates the smoothed real-time thermal coupling factor. Real-time correction of D in the feedforward decoupling matrix 21 (k) parameter (instantaneous value correction);

[0156] The slow loop uses minute-level time granularity and employs FF-RLS to identify and update the system parameter vector online. The reference value D of the feedforward decoupling matrix is ​​synchronously corrected. 21,0 (Base value correction).

[0157] The fast and slow loops operate in parallel without interfering with each other, working together to ensure that the feedforward decoupling matrix remains in the optimal state at different time scales.

[0158] Preferably, the system further includes: a first temperature sensor disposed at the atomic gas chamber for collecting the gas chamber temperature; and a second temperature sensor disposed at the laser for collecting the laser temperature. Preferably, the gas chamber heater is a non-magnetic heating element disposed on the outer wall of the atomic gas chamber; and the laser heater is a TEC cooling element disposed on the laser module.

[0159] like Figure 4 As shown, the system of the present invention includes the following core modules:

[0160] 1) System modeling and virtual controller module: Establish the system model and generate virtual control variables V1 and V2;

[0161] 2) Feedforward decoupling module, which converts virtual control quantities into actual control output quantities U1 and U2;

[0162] 3) Short-timescale dynamic correction module (fast loop): Calculates dynamic thermal coupling factor based on time delay compensation, using the sampling period as the time granularity. And fine-tune the coupling parameters online;

[0163] 4) Long-term online identification module (slow loop): The system parameter vector is updated online using FF-RLS with a time granularity of minutes. And correct the baseline value of the decoupling matrix.

[0164] exist Figure 4 In the diagram, solid arrows represent the main control signal flow, dashed arrows represent the parameter update / correction signal flow, and fast loops and slow loops are distinguished by different colors to reflect their time scale differences.

[0165] like Figure 5 As shown, the horizontal axis represents time, and the vertical axis represents the laser temperature deviation. The dashed line represents the traditional PID scheme: under a +5℃ step disturbance in the gas chamber, the peak temperature deviation of the laser is about 0.25℃, exhibiting significant overshoot and oscillation; the solid line represents the scheme of this invention: the peak deviation is only 0.028℃ (about 1 / 9 of the traditional PID), with no overshoot and oscillation, and the settling time is shortened to less than 2 minutes. In other words, when a ±5℃ disturbance is applied to the atomic gas chamber, this invention can still effectively suppress laser temperature fluctuations.

[0166] Example 2

[0167] This embodiment provides a thermally coupled dynamic adaptive decoupling control method for a micro atomic magnetometer, applied to a SERF atomic magnetometer, wherein the atomic gas chamber is set to a temperature of 160°C, the laser is set to a temperature of 65°C, and the distance between the two is 3 mm.

[0168] Step 1: In the system initialization phase, a fixed step power excitation signal is applied to the atomic gas chamber heater and the laser heater respectively, and the dynamic temperature response data is collected synchronously. The least squares method is used to fit the FOPDT model, and the model parameters of each channel are identified as shown in Table 1.

[0169] Table 1 Thermal Coupling Identification Parameters

[0170]

[0171] As can be seen from Table 1, τ 21 =35s is much greater than τ 22 =3s, K 21 =1.8 is much greater than K 12 =0.25, which reflects the asymmetric thermal coupling characteristics.

[0172] Step 2: Based on the system model, design the cross-feedforward decoupling matrix D(s);

[0173] Step 3: Let the sampling period Δt = 1s, then d 21 =round(35 / 1)=35, the short-time scale time granularity is 1 second. Under the short-time scale (fast loop), the dynamic thermal coupling factor is calculated in real time with a time granularity of 1 second, and smoothing is performed using a sliding window length N=10, and based on... Real-time correction of D in the feedforward decoupling matrix 21 parameter.

[0174] Step 4: On a long time scale (slow loop), initiate FF-RLS online identification every 60 seconds (minutes). The time granularity of the long time scale is 60 seconds, which is 60 times the time granularity of the short time scale. The forgetting factor λ = 0.99, update the system parameter vector, and synchronously correct the baseline value D of the decoupling matrix. 21,0 .

[0175] Step 5: Convert the virtual control quantities V1 and V2 into actual heater control output quantities U1 and U2 through the decoupling matrix, and drive the atomic gas chamber heater and laser heater to work.

[0176] Tests showed that when a +5°C step disturbance was applied to the gas chamber, the laser temperature fluctuation was 0.028°C; the system stabilization time (entering the ±0.05°C error band) was 1.5 minutes; and the temperature control error was 0.021°C after 72 hours of continuous operation.

[0177] In the above embodiments, the method and system designed in this invention have several significant technical advantages compared to the traditional PID single-loop temperature control scheme. The specific technical effects are as follows:

[0178] 1) Achieving ultra-high precision steady-state temperature control. This invention significantly suppresses bidirectional thermal crosstalk interference through MIMO precise thermal coupling modeling and feedforward decoupling algorithms. Under extreme conditions where the gas chamber generates a large temperature fluctuation of ±5℃, the laser temperature fluctuation can be controlled within 0.03℃, and the system's steady-state temperature control accuracy can reach ±0.01℃, effectively locking the laser's operating wavelength and fundamentally improving the long-term magnetic field measurement accuracy and data stability of the atomic magnetometer; specific test results are shown in Table 2:

[0179] Table 2 Disturbance Test Results

[0180]

[0181] 2) Significantly reduces system startup lock-up time. During the equipment startup phase, relying on feedforward decoupling combined with historical control quantity advance prediction compensation technology, temperature disturbances caused by thermal shock in the gas chamber can be addressed in advance, significantly suppressing laser temperature overshoot. Experimental tests show that this invention can shorten the system steady-state lock-up time from more than 20 minutes in traditional PID control to 1.5 minutes, greatly improving the equipment's rapid response capability. Specific test results are shown in Table 3:

[0182] Table 3 Startup Performance Comparison

[0183]

[0184] The stabilization time refers to the time it takes for the temperature to stabilize after entering the ±1℃ error band and not exceed the required range; the ±0.05℃ time is the time it takes to first enter the error band.

[0185] 3) Ensuring stable operation throughout the equipment's entire lifecycle. Through the FF-RLS online adaptive iterative algorithm, the system can detect thermodynamic parameter drift in real time, dynamically update the control model and decoupling matrix, effectively adapting to long-term operating condition changes such as component aging and vacuum degradation, completely solving problems such as traditional temperature control model mismatch, temperature oscillation, and temperature drift, significantly extending the equipment's high-precision stable operation cycle. Through a dynamic thermal coupling factor online correction mechanism, this invention can detect changes in thermal coupling strength in real time and automatically adjust the decoupling compensation capability. In long-term operation experiments, even if the thermal coupling gain changes by more than 15%, the system can still maintain the decoupling control effect, and the laser temperature fluctuation increase does not exceed 0.005℃. Specific test results are shown in Table 4:

[0186] Table 4 Results of long-term operation

[0187]

[0188] 4) Significantly reduces the impact of thermal disturbances. This invention accurately cancels coupling interference through a decoupling matrix, and with the help of an asymmetric time delay advance compensation strategy, the laser temperature overshoot is reduced by about 70%, and the impact of thermal disturbances on temperature control accuracy is reduced by 5 to 10 times. It can maintain high-precision constant temperature control under various steady-state and dynamic disturbance conditions, and has extremely strong environmental adaptability and anti-interference ability.

[0189] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A thermally coupled dynamic adaptive decoupling control method for a micro atomic magnetometer, characterized in that, include: A thermal coupling transfer function model between the atomic gas chamber and the laser is established. The thermal coupling transfer function model includes the thermal temperature control main channel transfer function of the atomic gas chamber, the thermal temperature control main channel transfer function of the laser, the thermal disturbance coupling channel transfer function of the atomic gas chamber to the laser, and the thermal disturbance coupling channel transfer function of the laser to the atomic gas chamber. Based on the thermal coupling transfer function model, a feedforward decoupling matrix is ​​designed to convert the virtual control quantity into the actual heater control output quantity. On a short timescale, with the system sampling period Δt as the time granularity, the dynamic thermal coupling factor is calculated based on the pure time delay compensation of the thermal disturbance channel of the laser by the atomic gas cell, and the feedforward decoupling matrix is ​​corrected in real time to suppress transient thermal disturbances. On a long time scale, with a time granularity of 30 to 300 seconds, the recursive least squares algorithm with a forgetting factor is used to update the system thermodynamic parameters online and simultaneously correct the reference value of the feedforward decoupling matrix. The real-time correction of the dynamic thermal coupling factor at the short time scale and the online update of the recursive least squares algorithm at the long time scale run in parallel with different time granularities, and both act on the same feedforward decoupling matrix. The time granularity of the long time scale is at least 60 times that of the time granularity of the short time scale.

2. The method according to claim 1, characterized in that: The calculation of the dynamic thermal coupling factor and real-time correction of the feedforward decoupling matrix based on the pure time delay compensation of the laser thermal perturbation channel in the atomic gas cell at a short time scale includes: Calculate real-time thermal coupling strength index : ； in, This represents the temperature change of the laser at time k. This represents the temperature change of the atomic gas chamber after thermal coupling delay compensation. Thermal coupling channel pure time delay The corresponding number of discrete sampling steps; The system sampling period; Let K be the real-time temperature of the laser at time k. for Real-time temperature of the atomic gas chamber; The smoothed real-time thermal coupling factor is calculated using a sliding window integration method with a dead zone. : ； in, The length of the sliding window; A regularization factor is used to prevent the denominator from approaching zero; The decoupling compensation parameters in the feedforward decoupling matrix are corrected in real time based on the smoothed real-time thermal coupling factor. ； in: The nominal decoupling parameters identified during system initialization; This is the reference thermal coupling factor under rated operating conditions; To adaptively adjust the gain coefficient; These are the decoupling compensation parameters updated in real time.

3. The method according to claim 1, characterized in that: In the thermal coupling transfer function model, the pure time delay τ of the atomic gas cell on the thermal perturbation coupling channel of the laser is... 21 The pure time delay τ of the laser master control channel is greater than 22 Furthermore, the steady-state gain K of the atomic gas cell on the thermal perturbation coupling channel of the laser is... 21 The steady-state gain K of the laser on the thermal disturbance coupling channel of the gas chamber is greater than that of the gas chamber. 12 .

4. The method according to claim 1, characterized in that: The feedforward decoupling matrix is: ； The formula for converting virtual control quantity to actual heater control output quantity is: ； in, This is a virtual control quantity for the atomic gas chamber. This is a virtual control variable for the laser; This is the actual control output of the atomic gas chamber. This refers to the actual control output of the laser heater; G 11 (s) is the transfer function of the main channel for room temperature control of atomic gas, G 22 (s) is the transfer function of the main channel for laser temperature control, G 12 (s) is the transfer function of the laser to the thermal perturbation coupling channel of the atomic gas cell, G 21 (s) is the transfer function of the thermal perturbation coupling channel of the atomic gas cell to the laser.

5. The method according to claim 1, characterized in that: The feedforward decoupling matrix is ​​discretized through a bilinear transformation and combined with a historical control quantity buffer FIFO queue determined by the time delay difference to achieve advance hedging control in the discrete domain; the time delay difference includes the pure time τ of the gas cell to the laser thermal disturbance channel. 21 With the pure time delay τ of the laser master control channel 22 difference.

6. The method according to claim 1, characterized in that, The recursive least squares algorithm with a forgetting factor operates at a fixed time interval T. RLS Once started, the regression vector contains temperature measurements and control output values ​​from multiple sampling times prior to the current time. The recursive least squares algorithm with a forgetting factor updates the system parameter vector containing the steady-state gain and time constant of each channel online.

7. A thermally coupled dynamic adaptive decoupling control system for a micro atomic magnetometer, used to implement the method according to any one of claims 1 to 6, characterized in that, include: A temperature control loop for atomic gas chambers is used to collect the temperature of the atomic gas chambers and generate virtual control quantities for the atomic gas chambers based on the deviation between the setpoint and the measured value. The laser temperature control loop is used to collect the laser temperature and generate a virtual control quantity for the laser based on the deviation between the temperature setpoint and the measured value. The feedforward decoupling module is used to design a feedforward decoupling matrix based on the thermal coupling transfer function model between the atomic gas chamber and the laser, and to convert the virtual control quantities of the atomic gas chamber and the laser into the actual control output quantities of the atomic gas chamber heater and the laser heater. The time delay compensation module is used to determine the depth of the historical control quantity buffer FIFO queue based on the asymmetric time delay difference between the laser thermal disturbance channel and the laser main control channel in the atomic gas cell, so as to realize the advance hedging control in the discrete domain. The short-timescale dynamic correction module is used to calculate the dynamic thermal coupling factor and correct the feedforward decoupling matrix in real time, based on the pure time delay compensation of the thermal disturbance channel of the laser by the atomic gas cell, with the system sampling period Δt as the time granularity. The long-term online identification and update module is used to update the system thermodynamic parameters online with a time granularity of 30 to 300 seconds using a recursive least squares algorithm with a forgetting factor, and simultaneously correct the reference value of the feedforward decoupling matrix. The heating drive module is used to drive the atomic gas chamber heater and the laser heater according to the actual control output.

8. The system according to claim 7, characterized in that, The atomic gas room temperature control loop includes: The first temperature sensor is located in the atomic gas chamber and is used to collect the real-time temperature of the atomic gas chamber. The first virtual controller is used to generate virtual control quantities for the atomic gas chamber based on the deviation between the setpoint temperature of the atomic gas chamber and the real-time temperature of the atomic gas chamber.

9. The system according to claim 7, characterized in that, The laser temperature control circuit includes: The second temperature sensor is located at the laser and is used to collect the real-time temperature of the laser. The second virtual controller is used to generate a virtual control quantity for the laser based on the deviation between the laser temperature setpoint and the real-time temperature of the laser.

10. The system according to claim 7, characterized in that, The delay compensation module determines the buffer depth of the FIFO queue based on the delay difference. The FIFO queue is used to store historical control quantities. When performing discrete domain decoupling operations, the feedforward decoupling module reads historical control quantities from the FIFO queue to achieve advance hedging control.

Images