A state-aware bumpless transfer control method for a heat-pipe-cooled dual-loop system

Abstract

This invention relates to the field of power supply and distribution system technology, and discloses a state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system. The method includes: acquiring the temperature of the hot nodes and the steady-state integral operator values ​​representing the heat dissipation state of the power load in the main heat dissipation distribution branch, and simultaneously monitoring the cold-state temperature response data of the standby branch; constructing a dynamic thermal resistance topology model and substituting the steady-state integral operator values ​​to invert and calculate the initial integral value; resetting the integral operator of the standby branch to the initial integral value when switching is triggered, and outputting a control command superimposed with a micro-amplitude flutter signal to drive the thermal power regulation execution unit. This invention reconstructs the switching logic to the functional unit logic domain of automatic control, collaboratively eliminating mechanical execution dead zones and latent heat hysteresis of phase transitions, suppressing transient thermal pulsations in the power distribution environment, and ensuring the thermal safety of the power supply link for high-power loads.

Description

Technical Field

[0001] This invention belongs to the technical field of power supply and distribution systems, and particularly relates to a state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system. Background Technology

[0002] High-power electrical equipment continuously releases extremely high heat flux density during operation. The dual-loop redundant system of heat pipe cooling constitutes the basic physical architecture to ensure its continuous and stable power supply. The industry generally adopts the strategy of alternating operation of the main and backup dual loops to maintain the thermoelectric balance of electrical components. In the automatic control system with redundant design, the smooth switching of each adjustment function unit is the core to ensure the stability of the controlled object. However, for complex controlled systems with nonlinear characteristics, the existing control logic mostly adopts rigid switching triggered by threshold or proportional weighted scheduling with preset clock cycles. The above conventional scheduling logic treats the handover of control as an independent discrete-time event. However, in the actual physical operation, the gas-liquid phase change process of the working fluid inside the heat pipe has thermal inertia and nonlinear hysteresis characteristics. When the inactive loop is in a cold resting environment for a long time, the internal fluid quiescence forms a physical awakening barrier. At the same time, the integral operator of the controller lacks real flow field state feedback during standby, which will inevitably produce open-loop saturation and asynchronous cumulative deviation. If the control weight is forcibly transferred according to the solidified time domain curve, the execution pace of the control command will be out of sync with the actual latent heat absorption progress of the energy domain inside the pipeline.

[0003] Analysis of existing control methods reveals the following main shortcomings: 1. Static logic based on time-series scheduling cannot dynamically respond to the intensity of actual thermal load impacts. When the electrical network encounters transient high thermal load conditions, the fixed switching speed causes the heat dissipation flow field establishment rate of the inactive loop to lag behind the heat accumulation rate of the controlled object, resulting in a transient gap in the total cooling capacity; 2. The control signal outputs a large step command at the moment of takeover, causing high-frequency mechanical impacts in the regulating valve. The pipeline pulse pressure caused by the dead zone crossing process of this mechanical action is directly converted into physical thermal disturbances in the precision electrical equipment; 3. The heat transfer lag state of the cold standby loop lacks a pre-detection and state phase alignment mechanism before switching. Forced takeover causes a secondary overshoot in the temperature of the controlled end. This severe temperature fluctuation causes a severe drift in the junction temperature and electrical impedance of the power semiconductor device, thereby triggering the safety degradation or power outage protection of the power supply and distribution system.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a dynamic control mechanism to sense the real-time evolution of the physical thermal resistance state inside the dual loop, transform the cold physical barrier of the non-activated loop into an adaptive control handover parameter, eliminate the phase change hysteresis and heat dissipation trap in the wake-up process, realize the smooth and uninterrupted takeover of the redundant system under extreme thermal shock conditions, and ensure the continuous stability of the output power of the power supply and distribution network of high-power equipment. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system, comprising the following steps:

[0006] The system acquires real-time hot node temperature data representing the heat dissipation state of the power load in the main heat dissipation distribution branch that is in active operation, opening feedback data of the current thermal power regulation execution unit in the main heat dissipation distribution branch, and steady-state integral operator value of the main heat dissipation distribution branch controller. Simultaneously, it acquires cold medium end temperature response data representing the initial thermal energy storage boundary of the standby heat dissipation distribution branch to be switched.

[0007] A dynamic thermal resistance topology model is constructed based on the deviation gradient between real-time hot node temperature data and the preset power load target control temperature. The steady-state integral operator value is substituted into the dynamic thermal resistance topology model for inverse inversion to calculate the integral preset initial value of the backup heat dissipation distribution branch.

[0008] At the timing node of receiving the switching command, the integral accumulation logic of the backup heat dissipation distribution branch controller is blocked, the state of the integral operator of the backup heat dissipation distribution branch controller is reset to the integral preset initial value, the opening control command is generated based on the integral preset initial value, and a micro-amplitude flutter drive signal with preset frequency characteristics is generated.

[0009] The micro-amplitude flutter drive signal is superimposed on the opening control command to generate a composite drive command, and the composite drive command is output to the thermal power regulation execution unit of the backup heat dissipation distribution branch to set the initial current flow cross-sectional area of ​​the thermal power regulation execution unit.

[0010] Preferably, in the step of calculating the integral preset initial value of the backup heat dissipation distribution branch, the integral preset initial value... The calculation follows the following quantification relationship: ,in, This is the thermal resistance asymmetry correction coefficient determined based on the dynamic thermal resistance topology model. The steady-state integral operator value of the main heat dissipation distribution branch controller, where α is the preset latent heat compensation gain factor of the medium. To preset the target control temperature for the power load, For real-time hot node temperature data, To calculate the start time, This refers to the timing node that receives the switching instruction.

[0011] Preferably, the step of acquiring the cold-state medium end temperature response data characterizing the initial thermal energy storage boundary of the backup heat dissipation distribution branch to be switched includes: before receiving the timing node of the switching command, sending a test pulse excitation signal to the thermal power regulation execution unit of the backup heat dissipation distribution branch, wherein the amplitude of the test pulse excitation signal is less than a preset static seal maintenance threshold; acquiring the medium end temperature change rate of the backup heat dissipation distribution branch after receiving the test pulse excitation signal; extracting the delay time feature and amplitude attenuation ratio feature from the medium end temperature change rate, and multiplying the delay time feature and amplitude attenuation ratio feature to generate a phase change damping parameter; using the phase change damping parameter as a proportional attenuation factor to limit the output ramp slope of the integral preset initial value.

[0012] Preferably, the step of generating a micro-amplitude flutter drive signal with preset frequency characteristics includes: reading static friction threshold data and flow sensitivity dead zone data of the thermal power regulation execution unit from a preset memory; generating a sine wave disturbance characteristic quantity, limiting the amplitude of the disturbance characteristic quantity to be between the static friction threshold data and the flow sensitivity dead zone data, and limiting the frequency of the disturbance characteristic quantity to be greater than the preset thermal inertia cutoff frequency; and using the disturbance characteristic quantity as the micro-amplitude flutter drive signal.

[0013] Preferably, the step of calculating the integral preset initial value of the backup heat dissipation distribution branch includes: based on the pre-stored electro-thermal conversion efficiency mapping table, retrieving the corresponding efficiency ratio factor according to the deviation gradient; using the efficiency ratio factor to perform product weighting processing on the steady-state integral operator value to obtain the basic control quantity; multiplying the basic control quantity with the nonlinear correction coefficient output by the dynamic thermal resistance topology model to obtain the integral preset initial value, and using the integral preset initial value as the output limiting boundary of the opening control command.

[0014] Preferably, after the step of outputting the synthetic drive command to the thermal power regulation execution unit of the backup heat dissipation distribution branch, the method further includes: calculating the current heat dissipation rate based on the real-time collected inlet and outlet water temperature difference and flow feedback data of the backup heat dissipation distribution branch, and calculating the deviation value between the current heat dissipation rate and the historical heat dissipation rate of the main heat dissipation distribution branch before switching; when the deviation value is greater than the preset safety deviation margin, the proportional-integral-derivative calculation logic of the bypass backup heat dissipation distribution branch controller is used to generate a step-type emergency opening correction command with the target opening set to the maximum value; and outputting the step-type emergency opening correction command to the thermal power regulation execution unit.

[0015] Preferably, the steps for constructing a dynamic thermal resistance topology model include: acquiring the liquid supply temperature data at the inlet of the main heat dissipation distribution branch, calculating the temperature difference between the liquid supply temperature data and the real-time thermal node temperature data, multiplying the temperature difference by a preset specific heat capacity parameter to determine the real-time sensible heat exchange rate; dividing the real-time sensible heat exchange rate by the effective flow area corresponding to the opening feedback data to calculate the medium phase change heat transfer coefficient; inputting the medium phase change heat transfer coefficient into a preset system thermal resistance network structure, extracting the thermal resistance node with the largest value as the key thermal resistance node; and generating a dynamic thermal resistance topology model based on the values ​​of the key thermal resistance nodes.

[0016] Preferably, the method further includes a preheating step performed before the timing node of receiving the switching command: calculating the fluctuation variance of the steady-state integral operator value within a preset time window; when the fluctuation variance is greater than a preset instability threshold, outputting a preheating drive command to the electrothermal power compensation unit of the backup heat dissipation distribution branch; setting the output power of the electrothermal power compensation unit based on the preheating drive command, so that the pipe wall temperature of the backup heat dissipation distribution branch rises to a preset gas-liquid two-phase coexistence temperature range.

[0017] Preferably, the method further includes: acquiring a load prediction signal containing the computing load change trend within a future preset time window; when the computing load change trend is monotonically increasing, superimposing a positive bias on the integral preset initial value; and converting the value after superimposing the positive bias into the initial displacement driving amount of the opening control command.

[0018] Compared with existing technologies, the state-aware, disturbance-free switching control method for the heat pipe cooling dual-loop system of the present invention has the following advantages:

[0019] 1. In the state perception of a dual-loop heat pipe cooling system, this invention addresses the problem of inconsistent initial states faced by redundant control systems during the initial switching of functional units. It constructs a cross-domain state perception control logic. By acquiring the first temperature difference during the steady-state operation of the active loop and the second temperature difference under the background cold environment of the inactive loop, and calculating the thermal resistance distortion compensation coefficient, this coefficient is multiplied by the steady-state integral operator value and directly written into the integral register of the inactive loop controller. This accurately transforms the physical characteristics of the cold-state wake-up barrier of the inactive loop into the initial bias kinetic energy in the integral register. This ensures that the control commands output by the system naturally contain the additional starting energy required to overcome the phase change hysteresis of the working fluid under specific heat loads, thereby effectively eliminating the heat capacity gap in the control system during the initial redundant switching and maintaining stable output of electrical parameters at the load end.

[0020] 2. Conventional weighted scheduling methods based on static clock cycles cannot perceive the true intensity of thermal load impact. When the electrical network encounters a sudden high heat flow condition, the fixed instruction timing will cause the cooling output to be misaligned with the actual heating process, thus triggering the overload protection of electrical equipment. This invention uses a differential operation unit to continuously extract the evolution slope of the temperature process variable of the controlled object. When the evolution slope approaches zero and the temperature process variable converges to the target set value, the system synchronously adjusts the weighted attenuation parameter of the active loop and the weighted gain parameter of the non-active loop. This mechanism makes the handover progress of control completely controlled by the physical dissipation process of the closed-loop temperature deviation, enabling the system to dynamically adjust the execution speed of control instructions according to the current changes in the heating intensity of electrical equipment. This ensures that the total cooling power envelope synthesized by the dual loop remains constant throughout the switching process, preventing power outage faults caused by the accumulation of local hot spots, and enhancing the adaptability of the power distribution network under extreme dynamic thermal loads and the safety of continuous power supply.

[0021] 3. The inherent static friction dead zone of physical regulating valves causes discontinuous action jumps in control commands at the end of switching periods with extremely small rates of change. This mechanical fluid jamming leads to pipeline pressure shocks and thermal pulsations, directly disrupting the physical stability of the precision power supply environment. This invention generates a sinusoidal disturbance characteristic quantity with a specific frequency and amplitude between static friction and flow sensitivity threshold when the time-varying weight function enters the nonlinear critical region. This quantity is then positively superimposed onto the undisturbed control command sequence. This high-frequency micro-amplitude excitation keeps the valve core of the actuator in a dynamically micro-amplitude suspended state. By using the mechanism of dynamic friction replacing static friction, the minimal linear increment of the weight function is transformed into a continuous and realistic fluid ratio change. This eliminates transient thermal disturbances in the power distribution environment caused by mechanical hysteresis of the actuator. Based on universal hardware, it achieves smooth pipeline flow control and ensures the physical connection safety of the power supply link for high-performance electrical equipment. Attached Figure Description

[0022] Figure 1 This is a flowchart of the control logic for the state-aware, disturbance-free switching of the dual-loop heat pipe cooling system of the present invention.

[0023] Figure 2 This is the core technical element and logical architecture diagram of the disturbance-free switching control method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0025] A state-aware, disturbance-free switching control method for a heat pipe-cooled dual-loop system includes the following steps:

[0026] The system acquires real-time hot node temperature data representing the heat dissipation state of the power load in the main heat dissipation distribution branch that is in active operation, opening feedback data of the current thermal power regulation execution unit in the main heat dissipation distribution branch, and steady-state integral operator value of the main heat dissipation distribution branch controller. Simultaneously, it acquires cold medium end temperature response data representing the initial thermal energy storage boundary of the standby heat dissipation distribution branch to be switched.

[0027] A dynamic thermal resistance topology model is constructed based on the deviation gradient between real-time hot node temperature data and the preset power load target control temperature. The steady-state integral operator value is substituted into the dynamic thermal resistance topology model for inverse inversion to calculate the integral preset initial value of the backup heat dissipation distribution branch.

[0028] At the timing node of receiving the switching command, the integral accumulation logic of the backup heat dissipation distribution branch controller is blocked, the state of the integral operator of the backup heat dissipation distribution branch controller is reset to the integral preset initial value, the opening control command is generated based on the integral preset initial value, and a micro-amplitude flutter drive signal with preset frequency characteristics is generated.

[0029] The micro-amplitude flutter drive signal is superimposed on the opening control command to generate a composite drive command, and the composite drive command is output to the thermal power regulation execution unit of the backup heat dissipation distribution branch to set the initial current flow cross-sectional area of ​​the thermal power regulation execution unit.

[0030] Preferably, in the step of calculating the integral preset initial value of the backup heat dissipation distribution branch, the integral preset initial value... The calculation follows the following quantification relationship: ,in, This is the thermal resistance asymmetry correction coefficient determined based on the dynamic thermal resistance topology model. The steady-state integral operator value of the main heat dissipation distribution branch controller, where α is the preset latent heat compensation gain factor of the medium. To preset the target control temperature for the power load, For real-time hot node temperature data, To calculate the start time, This refers to the timing node that receives the switching instruction.

[0031] Preferably, the step of acquiring the cold-state medium end temperature response data characterizing the initial thermal energy storage boundary of the backup heat dissipation distribution branch to be switched includes: before receiving the timing node of the switching command, sending a test pulse excitation signal to the thermal power regulation execution unit of the backup heat dissipation distribution branch, wherein the amplitude of the test pulse excitation signal is less than a preset static seal maintenance threshold; acquiring the medium end temperature change rate of the backup heat dissipation distribution branch after receiving the test pulse excitation signal; extracting the delay time feature and amplitude attenuation ratio feature from the medium end temperature change rate, and multiplying the delay time feature and amplitude attenuation ratio feature to generate a phase change damping parameter; using the phase change damping parameter as a proportional attenuation factor to limit the output ramp slope of the integral preset initial value.

[0032] Preferably, the step of generating a micro-amplitude flutter drive signal with preset frequency characteristics includes: reading static friction threshold data and flow sensitivity dead zone data of the thermal power regulation execution unit from a preset memory; generating a sine wave disturbance characteristic quantity, limiting the amplitude of the disturbance characteristic quantity to be between the static friction threshold data and the flow sensitivity dead zone data, and limiting the frequency of the disturbance characteristic quantity to be greater than the preset thermal inertia cutoff frequency; and using the disturbance characteristic quantity as the micro-amplitude flutter drive signal.

[0033] Preferably, the step of calculating the integral preset initial value of the backup heat dissipation distribution branch includes: based on the pre-stored electro-thermal conversion efficiency mapping table, retrieving the corresponding efficiency ratio factor according to the deviation gradient; using the efficiency ratio factor to perform product weighting processing on the steady-state integral operator value to obtain the basic control quantity; multiplying the basic control quantity with the nonlinear correction coefficient output by the dynamic thermal resistance topology model to obtain the integral preset initial value, and using the integral preset initial value as the output limiting boundary of the opening control command.

[0034] Preferably, after the step of outputting the synthetic drive command to the thermal power regulation execution unit of the backup heat dissipation distribution branch, the method further includes: calculating the current heat dissipation rate based on the real-time collected inlet and outlet water temperature difference and flow feedback data of the backup heat dissipation distribution branch, and calculating the deviation value between the current heat dissipation rate and the historical heat dissipation rate of the main heat dissipation distribution branch before switching; when the deviation value is greater than the preset safety deviation margin, the proportional-integral-derivative calculation logic of the bypass backup heat dissipation distribution branch controller is used to generate a step-type emergency opening correction command with the target opening set to the maximum value; and outputting the step-type emergency opening correction command to the thermal power regulation execution unit.

[0035] Preferably, the steps for constructing a dynamic thermal resistance topology model include: acquiring the liquid supply temperature data at the inlet of the main heat dissipation distribution branch, calculating the temperature difference between the liquid supply temperature data and the real-time thermal node temperature data, multiplying the temperature difference by a preset specific heat capacity parameter to determine the real-time sensible heat exchange rate; dividing the real-time sensible heat exchange rate by the effective flow area corresponding to the opening feedback data to calculate the medium phase change heat transfer coefficient; inputting the medium phase change heat transfer coefficient into a preset system thermal resistance network structure, extracting the thermal resistance node with the largest value as the key thermal resistance node; and generating a dynamic thermal resistance topology model based on the values ​​of the key thermal resistance nodes.

[0036] Preferably, the method further includes a preheating step performed before the timing node of receiving the switching command: calculating the fluctuation variance of the steady-state integral operator value within a preset time window; when the fluctuation variance is greater than a preset instability threshold, outputting a preheating drive command to the electrothermal power compensation unit of the backup heat dissipation distribution branch; setting the output power of the electrothermal power compensation unit based on the preheating drive command, so that the pipe wall temperature of the backup heat dissipation distribution branch rises to a preset gas-liquid two-phase coexistence temperature range.

[0037] Preferably, the method further includes: acquiring a load prediction signal containing the computing load change trend within a future preset time window; when the computing load change trend is monotonically increasing, superimposing a positive bias on the integral preset initial value; and converting the value after superimposing the positive bias into the initial displacement driving amount of the opening control command.

[0038] Example 1: In a high-heat-flux-density equipment heat dissipation scenario undertaking high-concurrency computing power scheduling tasks, the main heat dissipation distribution branch is in an active operating state and the current computing power load shows a step-like monotonically increasing trend, while the backup heat dissipation distribution branch is in a cold, quiescent environment for a long time, with the internal heat transfer medium dormant, forming a physical awakening barrier. At this time, if the control weight is transferred according to the fixed time domain curve, the execution pace of the control command will lag behind the actual latent heat absorption progress of the energy domain inside the pipeline network, causing a transient trap in the total cooling capacity and a secondary overshoot of the controlled end temperature. Traditional regulation systems usually compensate for the heat transfer time lag by increasing the mechanical response speed of the actuator. However, the state-aware, non-disruptive switching control method transforms the redundant switching process into a state variable mapping operation within the control logic domain. By establishing a quantitative mathematical connection between the physical thermal resistance state and the control integral operator, the redundant switching is changed. The logical trigger boundary conditions enable the controller's output initial command to directly overcome the time hysteresis blind zone caused by the cold phase transition of the medium. When a load prediction signal containing the monotonically increasing trend of computing power load within a future preset time window is obtained, the system acquires the real-time hot node temperature data representing the power load heat dissipation state of the main heat dissipation distribution branch in active operation, the opening feedback data of the current thermal power adjustment execution unit in the main heat dissipation distribution branch, and the steady-state integral operator value of the main heat dissipation distribution branch controller. Simultaneously, it acquires the cold medium end temperature response data representing the initial thermal energy storage boundary of the standby heat dissipation distribution branch to be switched. Based on the deviation gradient between the real-time hot node temperature data and the preset power load target control temperature, the system constructs a dynamic thermal resistance topology model and extracts the thermal resistance node with the largest value as the key thermal resistance node to generate the thermal resistance asymmetry correction coefficient.

[0039] In this information exchange path, the discrete physical temperature data collected by the underlying sensing devices provides a quantitative input for the dynamic thermal resistance topology model, characterizing the degree of heat transfer hindrance in the flow field. The thermal resistance asymmetry correction coefficient generated by this topology model acts inversely on the control operator domain. The system substitutes the steady-state integral operator values ​​into the dynamic thermal resistance topology model for inverse inversion, calculating the integral preset initial value of the backup heat dissipation distribution branch. Specific initial values ​​for integration The calculation follows a quantization relationship. ,in This is the thermal resistance asymmetry correction coefficient determined based on the dynamic thermal resistance topology model. The steady-state integral operator value of the main heat dissipation distribution branch controller, where α is the preset latent heat compensation gain factor of the medium. To preset the target control temperature for the power load, For real-time hot node temperature data, To calculate the start time, At the timing node where the switching command is received, the system blocks the integral accumulation logic of the backup heat dissipation distribution branch controller and resets the state of the integral operator of the backup heat dissipation distribution branch controller to the calculated preset initial value of the integral. And the initial value is preset in this integral. Based on the above, a positive bias is superimposed to generate an opening control command. At the end drive stage of the thermal power regulation actuator, where the output opening control command is sent to the thermal power regulation actuator, there is a mechanical hysteresis contradiction between the small linear command increment and the inherent static friction dead zone of the physical regulating valve. The system reads the static friction threshold data and flow sensitivity dead zone data of the thermal power regulation actuator from a preset memory, generating a sinusoidal waveform disturbance characteristic. The amplitude of this disturbance characteristic is set between the static friction threshold data and the flow sensitivity dead zone data, and the frequency of the disturbance characteristic is set greater than the preset thermal inertia cutoff frequency. The system uses this disturbance characteristic as a micro-amplitude flutter drive signal and positively superimposes it onto the opening control command to generate a synthetic drive command. This high-frequency micro-amplitude excitation drives the control valve core of the actuator to cross the static friction threshold and maintain a dynamic micro-amplitude suspension state. The linear following characteristic of dynamic friction is used to eliminate the dead zone hindrance effect of static friction, enabling the control valve core to move beyond the static friction threshold and maintain a dynamic micro-amplitude suspension state based on the integral preset initial value. The incremental evolution of control commands is continuously realized as real changes in the cross-sectional area of ​​the fluid flow; the thermal power regulation execution unit of the backup heat dissipation distribution branch receives the synthesized drive command and establishes the initial displacement drive amount; the phase change wake-up response curve of the medium inside the backup heat dissipation distribution branch is phase aligned with the heat dissipation unloading curve of the main heat dissipation distribution branch; the temperature of the hot node in the power distribution environment is maintained within the fluctuation tolerance range of the preset power load target control temperature; and no transient thermal pulsation event triggering the power outage protection mechanism is generated in the power supply link.

[0040] Example 2: In the 50kW-level heat pipe cooling dual-loop hardware-in-the-loop physical test platform for high heat flux density equipment, both the main heat dissipation distribution branch and the backup heat dissipation distribution branch are equipped with electric regulating valves with a flow capacity of 150L / min. The underlying data is obtained jointly by a computational fluid dynamics simulation model based on the Navier-Stokes equations and a physical temperature sensor. The system introduces Gaussian white noise with a signal-to-noise ratio of 20dB and a power frequency harmonic interference signal with a frequency of 50Hz into the measurement channel to simulate the electromagnetic interference environment of the power distribution network. When setting the discrete sampling period parameter, a balance needs to be achieved between the real-time capture of phase change interface propulsion and the computational load of the digital signal processor. Based on the Nyquist sampling theorem and the frequency characteristics of the maximum gas-liquid interface propulsion rate of the working fluid under 50kW thermal shock, the sampling frequency is set to be more than four times the characteristic frequency to avoid high-frequency state information aliasing. The sampling period is determined to be 10ms.

[0041] The experiment constructed a multi-dimensional control system, establishing a control group 1 using a 5s clock cycle weight decay switching strategy, a control group 2 eliminating micro-amplitude flutter drive signals, and an experimental group using complete control logic. Gradient problem variables with thermal shock amplitudes of 10kW / s, 25kW / s, and 50kW / s were introduced into the test sequence to simulate the handover condition from the main heat dissipation distribution branch to the backup heat dissipation distribution branch. Intermediate state monitoring data of the experimental group under a 50kW / s thermal shock were extracted, showing that the controller collects real-time hot node temperature data with 20dB Gaussian white noise at a 10ms cycle and filters out high-frequency interference signals using a Kalman filter algorithm. The state transition weight of the Kalman filter is set to 0.02, the measurement noise weight is set to 0.85, the sampling frequency is fixed at 100Hz, and a circular buffer with a capacity of 10 samples is set inside the algorithm. A state update calculation is performed every 10 milliseconds. By weighting and correcting the prediction deviation and the measurement residual, it is ensured that while filtering out 50Hz power frequency harmonic interference, the physical phase lag of the signal is controlled within 15 milliseconds. This provides a smooth and low-latency temperature change rate input for subsequent differential operations. The deviation gradient at the node with the maximum thermal resistance is calculated to be 8.5℃ / s, and a thermal resistance asymmetry correction coefficient of 1.45 is generated. Combined with the extracted steady-state integral operator values ​​of the main heat dissipation distribution branch The value is 55.2, based on the quantitative relationship. Establish the initial value of the integral. The value is 82.6, where α is the preset latent heat compensation gain factor of the medium. To preset the target control temperature for the power load, For real-time hot node temperature data, To calculate the start time, For timing nodes that receive switching commands, the dynamic thermal resistance topology model quantifies the phase change wake-up resistance of the cold medium into a specific integral bias compensation amount, providing a driving reference for the actuator to cross the start-up dead zone.

[0042] Comparing the system terminal thermal response data under gradient operating conditions, the control group 1 experienced a second overshoot of 18.4℃ at the hot node temperature under a 50kW / s thermal shock, and a 4.2s cooling power gap occurred during the handover period. The experimental group relied on the reset integral preset initial value. A micro-diffraction drive signal with a frequency set to 15Hz and an amplitude set to 2.5% of the valve opening was used to drive the regulating valve core across the static friction threshold. The maximum overshoot of the thermal node temperature was 1.2℃ and the valve displacement curve showed continuous linear characteristics. In the out-of-range control group where the frequency of the micro-diffraction drive signal was swept from 0.5Hz to 25Hz, a nonlinear performance inflection point boundary was observed. When the frequency was lower than the preset thermal inertia cutoff frequency of 2.0Hz, the control mechanism could not overcome the static friction dead zone, causing the valve response lag time to surge to 1850ms. When the frequency exceeded the mechanical resonance upper limit of 20.0Hz, the actuator experienced structural oscillation and the flow sensitivity decreased by 45%. 15Hz was confirmed as the operating window that balances dead zone elimination and mechanical stability. The synergistic effect of integral preset compensation and dizziness disturbance supplemented the thermal capacity gap of the control system in the initial switching stage.

[0043] Example 3: During the hardware adaptation phase of the heat pipe cooling dual-loop system, the manufacturing tolerances of the regulating valves and the differences in friction resistance along the pipe network cause a reference drift in the integral preset value. The system performs physical calibration on the dynamic thermal resistance topology model and the frequency response boundary of the regulating valves. For the construction of the dynamic thermal resistance topology model, the system obtains the physical coordinate sequence of the temperature sensor array distributed in the pipe network space, and constructs a one-dimensional discrete spatial mapping matrix with the actual coolant flow path distance from the heating core to the cold plate return water end as the horizontal axis. The system injects a step heat load with an amplitude of 10% of the rated heat power into the main heat dissipation distribution branch, and simultaneously collects the real-time temperature of each discrete coordinate point in the spatial mapping matrix. The rate of change is calculated by dividing the difference in temperature change rates between adjacent coordinate points by the corresponding physical flow path distance to obtain the temperature deviation gradient array. In the dynamic thermal resistance topology model construction stage, a multi-order equivalent RC thermal network model is established based on the geometric configuration of the main and backup heat dissipation distribution branch pipes, pipe wall thickness, and medium flow parameters. Temperature sensors are used to acquire inlet and outlet temperature differences and hot node temperature data to calculate real-time sensible heat exchange and medium phase change heat transfer coefficient. The obtained data is used as input variables for the thermal resistance matrix. The eigenvalues ​​of the thermal resistance matrix are calculated, and node parameters representing the maximum heat transfer hindrance are extracted. These node parameters are then compared with a pre-stored reference thermal resistance benchmark to determine the thermal resistance asymmetry correction coefficient. , so that the integral is preset to an initial value The state mapping between the control logic domain and the physical energy domain is realized based on the difference in heat transfer efficiency of the current physical flow field.

[0044] After acquiring the temperature deviation gradient array, the system extracts the term with the largest absolute value from the array and identifies it as the critical thermal resistance node. Then, it calculates the ratio of the temperature deviation gradient at this critical thermal resistance node to a preset ideal heat transfer gradient benchmark value to generate a thermal resistance asymmetry correction coefficient. When determining the thermal inertia cutoff frequency of the micro-amplitude flutter drive signal, the system, with the backup heat dissipation distribution branch in a cold, quiescent state, sends a sinusoidal displacement detection signal with a frequency continuously increasing from 0.1Hz to 20.0Hz to the thermal power regulation execution unit. Simultaneously, the system collects the valve stem displacement feedback data, calculates the amplitude attenuation ratio and phase lag angle between the detection signal and the displacement feedback data, and records the excitation frequency when the amplitude attenuation ratio reaches the -3dB attenuation threshold and the phase lag angle reaches -45 degrees, setting this as the thermal inertia cutoff frequency. In determining the parameters of the micro-amplitude flutter drive signal, before the regulation mechanism operates, a continuously sweeping sinusoidal displacement command with a frequency range of 0.1Hz to 25Hz is sent to the thermal power regulation execution unit. The phase response of the valve stem displacement feedback sequence is collected, and the critical frequency when the amplitude attenuation reaches -3dB and the phase lag reaches -45° is set. Using the thermal inertia cutoff frequency as the input, a slowly varying drive current with a constant slope is input, and the micro-flow mutation point of the pipeline network is captured. The corresponding current command value is set as the static friction threshold data, and the initial current value when the flow enters the linear follow-up interval is set as the flow sensitivity dead zone data. The disturbance characteristic quantity forms a dynamic high-frequency micro-suspension effect at the valve core, so that the control command increment is converted into a continuous change in the fluid flow cross-sectional area. The discretization modeling and frequency domain sweep feature extraction process converts the degree of heat transfer hindrance and mechanical response limit into digital matrices and frequency parameters. The input of the dynamic thermal resistance topology model and the thermal inertia cutoff frequency constitutes the closed-loop condition of the integral preset initial value calculation logic. The frequency of the generated micro-amplitude flutter drive signal is positioned in a frequency band that avoids the mechanical attenuation dead zone and crosses the static friction threshold. The state-aware, disturbance-free switching control method outputs continuous fluid flow cross-sectional area changes on multiple batches of heat dissipation hardware.

[0045] Example 4: When the system faces initial commissioning conditions with different batches of cooling media or different chemical refrigerant injection volumes, the controller executes the quantitative filling procedure of the media latent heat compensation gain factor before the start of the entire network fluid circulation. The test terminal applies a constant amplitude simulated step heat load to the heated end of the backup heat dissipation distribution branch, driving the thermal power adjustment execution unit to the reference test opening. The temperature probe continuously collects the real-time temperature parameters of the working fluid in the loop during the process of temperature rise from the liquid phase sensible heat to the gas-liquid two-phase nucleus boiling region. The data processing unit extracts the temperature rise hysteresis time domain width corresponding to the phase change latent heat absorption stage and the dynamic temperature difference integral value during this period. This integration process is set to be received... Within the first 60 seconds after receiving the simulated load signal, the deviation between the real-time hot node temperature and the back temperature is calculated using a rectangular summation operation with a sampling interval of 10 milliseconds. The controller extracts 20% of the reference test opening value and divides it by the sum of the temperature difference integrals within the 60-second window. The calculated quotient is the latent heat compensation gain factor of the medium. This coefficient reflects the opening compensation intensity required per unit temperature rise integral and is used to eliminate the heat transfer gap in the backup circuit at the moment of switching. The controller extracts the reference test opening value and divides it by the dynamic temperature difference integral value. The calculation unit writes the calculated quotient into the memory as the initial measured value of the latent heat compensation gain factor α of the medium.

[0046] In setting the latent heat compensation gain factor α, during the heat dissipation circuit charging test, a constant amplitude simulated step heat load is applied to the heated end of the backup heat dissipation distribution branch. The time width from the sensible temperature rise of the working fluid in the tube to the nucleation stage of the gas-liquid two-phase system is recorded. The integral trajectory value of the temperature difference change over time during this process is extracted. The benchmark test opening degree of the thermal power regulation execution unit is divided by the integral trajectory value to determine the initial value of the latent heat compensation gain factor α, which is stored in the controller memory as a global call constant. This enables the conversion of the startup energy gain of the backup branch controller integral operator under different heat load gradients. The thermal control equipment deployed in the computer room performs thermal environment baseline calculations before undertaking high-concurrency computing loads. The calibration procedure involves the data acquisition terminal acquiring multi-point spatial temperature hash values ​​within a preset number of clock cycles while the power distribution space is in a static state. The filtering module removes data jump nodes from the multi-point spatial temperature hash values ​​and calculates the arithmetic mean of the remaining valid nodes. The system determines the arithmetic mean as the environmental background thermal offset. In the feedback adjustment stage, the real-time thermal node temperature data is subtracted from the environmental background thermal offset to generate calibration temperature parameters. The calibration temperature parameters, combined with the pre-calibrated latent heat compensation gain factor α of the medium, participate in the inverse calculation of the integral preset initial value, driving the instruction output node of the thermal power adjustment execution unit to align with the transient thermodynamic jump curve of the computing chip.

[0047] Example 5: Addressing the physical phenomenon of dead zone boundary drift caused by mechanical wear of physical regulating valves under long-cycle operation conditions, the system initiates an online polling recalibration procedure for mechanical hysteresis parameters before the backup heat dissipation distribution branch is put into grid connection. Under the premise of maintaining the cold resting boundary, the controller sends a slowly changing ramp drive current with a constant slope to the thermal power regulation execution unit. The fluid sensor array synchronously captures the micro-flow pulse change in the downstream pipeline. When the micro-flow pulse change exceeds the flow meter noise floor baseline, the system extracts the valve position command value corresponding to the slowly changing ramp drive current at the current instant and anchors it as static friction threshold data. The drive current continues to increase along the ramp function. The control logic unit continuously calculates the first derivative of the micro-flow pulse change with respect to the drive current. When the first derivative remains constant within a continuously preset number of sampling periods, it is determined that the valve has entered the linear regulation zone. The system writes the valve position command value at this time into the memory and updates it as flow sensitivity dead zone data. This procedure transforms the mechanical hysteresis characteristics at the physical level into discrete digital measurement coordinates.

[0048] To address the abnormal boundary of excitation failure caused by transient mechanical jamming during the injection of micro-amplitude flutter drive signals, the system incorporates online fault-tolerant compensation logic based on displacement divergence into the output link of the synthesized drive command. The displacement sensor transmits the control valve core displacement sequence of the actuator with a time resolution of 100μs. The differentiator calculates the real-time displacement change rate of this displacement sequence. When the synthesized drive command has been continuously output across a preset mechanical hysteresis time window and the real-time displacement change rate remains below a preset jamming judgment threshold, the multiplier multiplies the amplitude parameter of the original disturbance characteristic by an incremental step compensation. The coefficient γ generates an asymmetric impact disturbance characteristic quantity with local peak values, where γ is a dimensionless variable greater than 1 that increases linearly and monotonically with the duration of jamming. The system positively superimposes the updated asymmetric impact disturbance characteristic quantity into the opening control command until the real-time displacement change rate exceeds the jamming judgment threshold. The compensation logic channel is blocked and restored to the original amplitude. The closed-loop dynamic intervention mechanism based on physical displacement feedback offsets the influence of external mechanical interference on the phase change wake-up timing, enabling the backup heat dissipation distribution branch to establish the initial displacement driving quantity according to the predetermined integral preset initial value under mechanical wear conditions.

[0049] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A state-aware, disturbance-free switching control method for a heat pipe-cooled dual-loop system, characterized in that, Includes the following steps: The system acquires real-time hot node temperature data representing the heat dissipation state of the power load in the main heat dissipation distribution branch that is in active operation, opening feedback data of the current thermal power regulation execution unit in the main heat dissipation distribution branch, and steady-state integral operator value of the main heat dissipation distribution branch controller. Simultaneously, it acquires cold medium end temperature response data representing the initial thermal energy storage boundary of the standby heat dissipation distribution branch to be switched. A dynamic thermal resistance topology model is constructed based on the deviation gradient between real-time hot node temperature data and the preset power load target control temperature. The steady-state integral operator value is substituted into the dynamic thermal resistance topology model for inverse inversion to calculate the integral preset initial value of the backup heat dissipation distribution branch. At the timing node of receiving the switching command, the integral accumulation logic of the backup heat dissipation distribution branch controller is blocked, the state of the integral operator of the backup heat dissipation distribution branch controller is reset to the integral preset initial value, the opening control command is generated based on the integral preset initial value, and a micro-amplitude flutter drive signal with preset frequency characteristics is generated. The micro-amplitude flutter drive signal is superimposed on the opening control command to generate a composite drive command, and the composite drive command is output to the thermal power regulation execution unit of the backup heat dissipation distribution branch to set the initial current flow cross-sectional area of ​​the thermal power regulation execution unit.

2. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, In the step of calculating the integral preset initial value of the backup heat dissipation distribution branch, the integral preset initial value... The calculation follows the following quantification relationship: ,in, This is the thermal resistance asymmetry correction coefficient determined based on the dynamic thermal resistance topology model. The steady-state integral operator value of the main heat dissipation distribution branch controller, where α is the preset latent heat compensation gain factor of the medium. To preset the target control temperature for the power load, For real-time hot node temperature data, To calculate the start time, This refers to the timing node that receives the switching instruction.

3. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The steps for obtaining the cold-state medium-end temperature response data representing the initial thermal energy storage boundary of the backup heat dissipation distribution branch to be switched include: before receiving the timing node of the switching command, sending a test pulse excitation signal to the thermal power regulation execution unit of the backup heat dissipation distribution branch, wherein the amplitude of the test pulse excitation signal is less than a preset static seal maintenance threshold; collecting the medium-end temperature change rate of the backup heat dissipation distribution branch after receiving the test pulse excitation signal; extracting the delay time feature and amplitude attenuation ratio feature from the medium-end temperature change rate, and multiplying the delay time feature and amplitude attenuation ratio feature to generate a phase change damping parameter; using the phase change damping parameter as a proportional attenuation factor to limit the output ramp slope of the integral preset initial value.

4. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The steps for generating a micro-amplitude flutter drive signal with preset frequency characteristics include: reading static friction threshold data and flow sensitivity dead zone data of the thermal power regulation execution unit from a preset memory; generating a sinusoidal waveform disturbance characteristic quantity, limiting the amplitude of the disturbance characteristic quantity to be between the static friction threshold data and the flow sensitivity dead zone data, and limiting the frequency of the disturbance characteristic quantity to be greater than the preset thermal inertia cutoff frequency; and using the disturbance characteristic quantity as the micro-amplitude flutter drive signal.

5. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The steps for calculating the integral preset initial value of the backup heat dissipation distribution branch include: based on the pre-stored electro-thermal conversion efficiency mapping table, retrieving the corresponding efficiency ratio factor according to the deviation gradient; using the efficiency ratio factor to perform product weighting processing on the steady-state integral operator value to obtain the basic control quantity; multiplying the basic control quantity with the nonlinear correction coefficient output by the dynamic thermal resistance topology model to obtain the integral preset initial value, and using the integral preset initial value as the output limiting boundary of the opening control command.

6. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, After the step of outputting the synthetic drive command to the thermal power regulation execution unit of the backup heat dissipation distribution branch, the method further includes: calculating the current heat dissipation rate based on the real-time collected inlet and outlet water temperature difference and flow feedback data of the backup heat dissipation distribution branch, and calculating the deviation value between the current heat dissipation rate and the historical heat dissipation rate of the main heat dissipation distribution branch before switching; when the deviation value is greater than the preset safety deviation margin, the proportional-integral-derivative calculation logic of the bypass backup heat dissipation distribution branch controller is used to generate a step-type emergency opening correction command with the target opening set to the maximum value; and outputting the step-type emergency opening correction command to the thermal power regulation execution unit.

7. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The steps for constructing a dynamic thermal resistance topology model include: acquiring the liquid supply temperature data at the inlet of the main heat dissipation distribution branch, calculating the temperature difference between the liquid supply temperature data and the real-time thermal node temperature data, multiplying the temperature difference by the preset specific heat capacity parameter to determine the real-time sensible heat exchange rate; dividing the real-time sensible heat exchange rate by the effective flow area corresponding to the opening feedback data to calculate the medium phase change heat transfer coefficient; inputting the medium phase change heat transfer coefficient into the preset system thermal resistance network structure, extracting the thermal resistance node with the largest value as the key thermal resistance node; and generating a dynamic thermal resistance topology model based on the values ​​of the key thermal resistance nodes.

8. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The method also includes a preheating step performed before the timing node of receiving the switching command: calculating the fluctuation variance of the steady-state integral operator value within a preset time window; when the fluctuation variance is greater than a preset instability threshold, outputting a preheating drive command to the electrothermal power compensation unit of the backup heat dissipation distribution branch; setting the output power of the electrothermal power compensation unit based on the preheating drive command, so that the pipe wall temperature of the backup heat dissipation distribution branch rises to a preset gas-liquid two-phase coexistence temperature range.

9. The state-aware, disturbance-free switching control method for a heat pipe cooling dual-loop system according to claim 1, characterized in that, The method also includes: acquiring a load prediction signal containing the trend of computing load change within a future preset time window; when the computing load change trend is monotonically increasing, adding a positive bias amount on the basis of the integral preset initial value; and converting the value after adding the positive bias amount into the initial displacement driving amount of the opening control command.

Images