Multi-stage filtering control method and device for cooling fan

Abstract

The invention relates to the field of heat dissipation control technology, and discloses a multi-order filtering control method and device for a heat dissipation fan. The method comprises: collecting operating parameters of a multi-fan heat dissipation system to construct a dynamic state space model, and obtaining a system operating state equation and a temperature observation equation; inputting the system operating state equation and the temperature observation equation into an expectation maximization sliding window iterative extended Kalman filter to obtain a system state estimation value and a prediction error covariance matrix; performing a three-stage series design on the cutoff frequency of a low-pass filter, the passband range of a band-pass filter, and the coefficient of an adaptive filter to obtain a system multi-order filtering control model; based on the system multi-order filtering control model, calculating an airflow coupling influence coefficient matrix between multiple fans, and optimizing the total energy consumption, noise level, and heat dissipation efficiency to obtain a fan control optimization parameter set. The invention reduces energy consumption while ensuring heat dissipation efficiency, and realizes a dynamic balance between control units.

Description

Technical Field

[0001] The present invention relates to the field of heat dissipation control technology, and in particular to a multi-stage filtering control method and device for a heat dissipation fan. Background Art

[0002] As electronic devices develop towards high performance and miniaturization, heat dissipation issues are becoming increasingly prominent, and cooling fans are widely used as the most commonly used heat dissipation components. Traditional single control methods have problems such as response lag, high noise, and high energy consumption when facing complex heat dissipation requirements, making it difficult to meet the requirements of modern electronic devices for precise temperature control.

[0003] Although the multi-fan cooling system has stronger heat dissipation capacity, due to the airflow coupling effect between fans, the speed and airflow direction of each fan need to be coordinated and adjusted, which brings greater challenges to the control system. Traditional control methods assume that the state space and dynamic model remain unchanged, and cannot handle changes in control mode in real time, especially when the system needs to be dynamically adjusted under different heat dissipation requirements. In addition, the noise problem in the multi-fan system is more prominent. The PWM-based fan control strategy will generate noise during the speed regulation process, affecting the stability of the system and user experience. Each fan usually requires independent temperature tracking control, and due to the different heat dissipation requirements in different areas, control conflicts are prone to occur, resulting in a decrease in the overall performance of the system. Summary of the invention

[0004] The present invention provides a multi-stage filtering control method and device for a heat dissipation fan, which reduces energy consumption while ensuring heat dissipation efficiency and achieves a dynamic balance among control units.

[0005] In a first aspect, the present invention provides a multi-order filtering control method for a cooling fan, the multi-order filtering control method for a cooling fan comprising:

[0006] The operating parameters of the multi-fan cooling system are collected to construct a dynamic state space model, and the system operating state equation and temperature observation equation are obtained;

[0007] Input the system operation state equation and the temperature observation equation into the expectation maximization sliding window iterative extended Kalman filter, and obtain the system state estimation value and the prediction error covariance matrix by performing the conditional expectation calculation of step E and the parameter iterative update of step M;

[0008] According to the system state estimation value and the prediction error covariance matrix, a three-stage series design is performed on the cutoff frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter to obtain a system multi-order filtering control model;

[0009] Based on the multi-order filtering control model of the system, the airflow coupling influence coefficient matrix between multiple fans is calculated, and the total energy consumption, noise level and heat dissipation efficiency are optimized through the gradient iteration method to obtain the fan control optimization parameter set.

[0010] In a second aspect, the present invention provides a multi-order filter control device for a cooling fan, the multi-order filter control device for a cooling fan comprising:

[0011] The acquisition module is used to collect the operating parameters of the multi-fan cooling system to build a dynamic state space model and obtain the system operation state equation and temperature observation equation;

[0012] A calculation module is used to input the system operation state equation and the temperature observation equation into an expectation maximization sliding window iterative extended Kalman filter, and obtain a system state estimation value and a prediction error covariance matrix by performing the conditional expectation calculation of step E and the parameter iterative update of step M;

[0013] A design module is used to perform a three-stage series design on the cutoff frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter according to the system state estimation value and the prediction error covariance matrix, so as to obtain a system multi-order filtering control model;

[0014] The solution module is used to calculate the airflow coupling influence coefficient matrix between multiple fans based on the multi-order filtering control model of the system, and optimize the total energy consumption, noise level and heat dissipation efficiency through the gradient iteration method to obtain the fan control optimization parameter set.

[0015] A third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned cooling fan multi-order filtering control method.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned cooling fan multi-order filtering control method.

[0017] In the technical solution provided by the present invention, a dynamic state space model and an expectation-maximization sliding window iterative extended Kalman filter are constructed to accurately estimate the system state, effectively solving the problem of inaccurate state estimation in traditional control methods; a three-stage series filter design is adopted to realize hierarchical filtering of system noise, significantly reduce the fan operation noise, and improve the stability of the system; an airflow coupling influence coefficient matrix is ​​introduced, and multi-objective optimization is realized through Pareto optimization solution, thereby ensuring heat dissipation efficiency while reducing energy consumption; a distributed collaborative optimization strategy is adopted to solve the control conflict problem in a multi-fan system and achieve dynamic balance among the control units; a control strategy based on a multi-mode state machine is designed, and smooth switching between different operating modes is realized through a dynamic programming algorithm, thereby avoiding system oscillation during mode switching. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0019] Figure 1 A schematic diagram of the steps of a multi-order filtering control method for a heat dissipation fan in an embodiment of the present invention;

[0020] Figure 2 It is a structural schematic diagram of a multi-stage filtering control device for a heat dissipation fan in an embodiment of the present invention;

[0021] Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The embodiment of the present invention provides a multi-order filtering control method and device for a cooling fan. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the multi-order filtering control method of the heat dissipation fan in the embodiment of the present invention includes:

[0024] Step S1, collecting the operating parameters of the multi-fan cooling system to construct a dynamic state space model, and obtaining the system operating state equation and temperature observation equation;

[0025] It is understandable that the execution subject of the present invention may be a cooling fan multi-order filter control device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject.

[0026] Specifically, various operating parameters in the multi-fan cooling system are collected. Real-time temperature change information is obtained through multiple temperature sensors distributed in various key parts of the system. These temperature data are transmitted in the form of time series and sorted and grouped in the data buffer. The sorting process arranges the data according to the timestamp, and the grouping is aggregated according to the sensor position or other defined conditions to form a temperature observation vector representing the current state of the system. At the same time, the operating status of all fans in the system is monitored, especially the speed information of the fans. These speed signals are obtained through a high-speed sampling device to form a set of speed state vectors reflecting the dynamic characteristics of the fans. The speed state vector is input into the state transition calculation unit, and the state transition relationship reflecting the dynamic behavior of the system is extracted by analyzing the law of fan speed change, and the linkage law between the fans and the external control input is described. In order to consider the impact of vibration and airflow noise during fan operation on the system, vibration signals and noise signals are collected through special sensing equipment. After these complex signals are processed by frequency domain separation technology, a set of noise vectors representing the interference characteristics of the system are obtained. The noise vector is input into the control matrix calculation unit and combined with the input response characteristics of the system to generate a control matrix. The role of the control matrix is ​​to establish the connection between the external control input and the internal state of the system, ensuring that the control command can effectively adjust the system operation state in the presence of noise interference. Based on the above data processing, the temperature observation vector and the fan speed state vector are used to calculate and adjust the observation matrix through the optimization criterion. The role of the observation matrix is ​​to associate the actual observation value of the system with the internal state. The optimization process considers the noise interference factor in the observation signal, thereby improving the state estimation accuracy of the system. The state transfer relationship, control matrix, observation matrix and system noise vector are comprehensively calculated to construct the state space model of the system. The model includes the operating state equation and the temperature observation equation. The operating state equation describes the dynamic behavior of the system, such as the change law of the fan speed and how the system responds to external input; while the temperature observation equation defines the relationship between the observation value and the internal state of the system.

[0027] Step S2, inputting the system operation state equation and the temperature observation equation into the expectation maximization sliding window iterative extended Kalman filter, and obtaining the system state estimation value and the prediction error covariance matrix by executing the conditional expectation calculation of step E and the parameter iterative update of step M;

[0028] Specifically, the system operation state equation and temperature observation equation are input into the expectation maximization sliding window iterative extended Kalman filter, the length of the sliding window is set and the maximum number of iterations is determined. These parameters determine the width of the sliding window for data processing within the time range and the accuracy of iterative optimization. Based on the sliding window parameters, the system state vector is initialized and calculated. By assuming the initial state and utilizing prior information, the initial state vector and initial covariance matrix are obtained. In order to ensure the rationality of the initial state, the physical characteristics and historical data of the system are combined for analysis to provide a reliable starting point for the extended Kalman filter. The initial state vector and initial covariance matrix are input into the calculation unit of the E step. The core task of the E step is to calculate the conditional expected value of the state vector. By combining the state equation and the observation equation and using the theory of Kalman gain, the expected value of the state vector is obtained under the premise of minimizing the error. This process also requires calculating the covariance matrix of the expected value of the state vector to obtain the covariance matrix of the current state, which reflects the degree of uncertainty of the state vector. A smaller covariance means a higher estimation accuracy. The expected value and covariance matrix of the state vector are input into the calculation unit of the M step to update the system parameters. The goal of the M step is to update the state transfer matrix and the observation matrix through parameter iteration optimization. These two matrices describe the dynamic evolution law of the system state in time and the mapping relationship between the system observation data and the state. By continuously adjusting these two matrices, the filter can better adapt to the dynamic changes of the system and improve the accuracy of the state estimation. The covariance matrix of the system noise and the observation noise is calculated when the state transfer matrix and the observation matrix are updated. The noise covariance matrix describes the statistical characteristics of random interference and measurement errors in the system, and their calculation is based on the updated state transfer matrix and observation matrix. By accurately calculating the noise covariance matrix, the performance of the Kalman filter is optimized, the sensitivity to noise is reduced, and the robustness of the filter is improved. The expected value of the state vector and the noise covariance matrix are input into the prediction unit of the extended Kalman filter to calculate the state prediction value at the next moment. The prediction unit uses the updated state transfer matrix and noise covariance matrix to predict the future state of the system and obtain the estimated value of the system state, which reflects the best estimate of the system state at the next moment based on the current information. Based on the estimated value of the system state, the covariance of the prediction error is calculated to generate the prediction error covariance matrix, which describes the degree of uncertainty in the system state estimation.

[0029] Step S3, according to the system state estimation value and the prediction error covariance matrix, the cut-off frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter are designed in three stages in series to obtain a system multi-order filtering control model;

[0030] Specifically, according to the estimated value of the system state, the power spectral density analysis is performed on the spectrum of the system noise. By calculating the power distribution of the system noise signal at different frequencies, the frequency distribution characteristics of the noise are revealed. By analyzing the main range and energy concentration point of the noise frequency distribution, the optimal cutoff frequency of the low-pass filter is determined. According to the optimization calculation results, the transfer function of the low-pass filter is designed, which describes the attenuation characteristics of the filter for signals of different frequencies, thereby forming a first-order filter. The prediction error covariance matrix is ​​input into the bandpass filter design unit. The bandpass filter extracts the effective frequency band in the signal and suppresses the redundant frequency components. Based on the prediction error covariance matrix, the passband range of the bandpass filter is dynamically adjusted by using the frequency band optimization algorithm. The algorithm identifies the frequency components in the system signal that are of great significance to the control by analyzing the frequency characteristics in the covariance matrix, and optimizes the passband range of the bandpass filter accordingly, and designs the transfer function of the second-order filter. The transfer function of the first-order filter and the transfer function of the second-order filter are cascaded to obtain the cascaded transfer function, which simultaneously realizes the suppression of high-frequency noise and the extraction of the effective signal frequency band. The cascaded transfer function is input into the adaptive filter design unit. The adaptive filter dynamically adjusts the filter parameters according to the real-time data to adapt to the changing characteristics of the system. By adopting the minimum mean square error criterion, the design process of the adaptive filter continuously iterates and optimizes the filter coefficients to ensure that it can achieve the best filtering effect under different operating conditions and generate the transfer function of the third-order filter. According to the transfer function of the third-order filter, the filter response characteristics are analyzed in the frequency domain to reveal the gain and phase characteristics of the three-stage filter at different frequencies and generate the frequency response curve of the system. Through this curve, the comprehensive performance of the three-stage filter is evaluated, and on this basis, the stability of the series structure of the filter is analyzed. The stability analysis aims to ensure that the series structure of the filter does not cause unnecessary oscillation or nonlinear effects during operation, thereby optimizing the system transfer function. The system transfer function is combined with the system state estimate to construct a complete multi-order filter control model. The model comprehensively considers the different functions of low-pass, band-pass and adaptive filters, realizes hierarchical optimization processing of signals, and ensures efficient and stable operation of the system under various operating conditions.

[0031] Step S4: Based on the system multi-order filtering control model, the airflow coupling influence coefficient matrix between multiple fans is calculated, and the total energy consumption, noise level and heat dissipation efficiency are optimized by the gradient iteration method to obtain the fan control optimization parameter set.

[0032] Specifically, the system transfer function in the multi-order filter control model of the system is spatially decomposed, and the overall dynamic behavior of the system is analyzed into the independent airflow characteristics of each fan. Through the decomposition operation, the airflow characteristic function of each fan is obtained, which describes the influence characteristics of each fan on the airflow under different operating conditions. The airflow characteristic function is input into the airflow coupling analysis unit, and the interaction coefficient between fans is analyzed to obtain the interaction coefficient between fans. These coefficients reflect the linkage effect caused by airflow interference between fans. On this basis, according to the interaction coefficient between fans, the airflow velocity and pressure distribution of each fan are numerically calculated by fluid dynamics to obtain an accurate airflow coupling influence coefficient matrix, which contains the complex airflow interaction information between fans in the system. The airflow coupling influence coefficient matrix is ​​input into the energy consumption modeling unit, and the total energy consumption of the fan group under different working conditions is calculated by combining the power characteristic curve of each fan. The energy consumption modeling process requires the integral calculation of the power consumption when the fan is running to obtain the energy consumption target function. At the same time, based on the airflow coupling influence coefficient matrix, the aerodynamic noise and mechanical noise of the fan group are analyzed. This process includes acoustic characteristic modeling and calculation, and the noise target function is generated by weighted superposition of the characteristics of noise at different frequencies. The noise objective function describes the noise level of the system under different working conditions. The airflow coupling influence coefficient matrix is ​​input into the heat dissipation modeling unit, and the heat transfer and temperature distribution of the entire heat dissipation system are calculated through the thermodynamic transfer equation to obtain the heat dissipation efficiency objective function. The energy consumption objective function, the noise objective function and the heat dissipation efficiency objective function are combined to construct the Pareto objective function, which comprehensively considers the energy consumption, noise and heat dissipation performance of the system to ensure that the requirements of multiple objectives can be weighed during the optimization process. The Pareto objective function is input into the gradient optimization unit, and the non-dominated optimal solution sequence is obtained through iterative calculation of different weight combinations. These solutions are the optimal balance points of each objective under the current weights, representing the best performance combination of the multi-fan system under different operating conditions. The non-dominated optimal solution sequence is transformed by parameter mapping to generate a specific fan control optimization parameter set, which includes key parameters such as fan speed, airflow angle and power distribution.

[0033] The fan control optimization parameter set is input into the distributed computing unit, and the control parameters are decoupled. By decomposing the global optimization parameter set, an initial control parameter set adapted to each control unit is generated to reflect the initial operating state of each fan in the system as a whole. Data acquisition and neighborhood information exchange are performed on the temperature sensor signals of multiple control units to generate a temperature state information interaction matrix to describe the temperature distribution relationship and dynamic change characteristics between different control units in the system. Based on the temperature state information interaction matrix, the temperature coupling strength between control units is quantitatively analyzed. Through numerical calculation, the degree of temperature influence between different control units is clarified to provide a basis for building a distributed communication topology. By introducing a weighted network model, a topology reflecting the coupling relationship between control units is constructed according to the temperature coupling strength, and a temperature coupling constraint matrix is ​​generated. The temperature coupling constraint matrix is ​​input into the local controller. In the calculation process of the local controller, the fan speed and airflow direction of each control unit are constrained and optimized. The optimization process uses the calculation model of local control and combines the information in the temperature coupling constraint matrix to generate the local control optimization vector of each control unit. The local optimization vector reflects the optimal control parameters of each control unit under specific temperature and load conditions. According to the local control optimization vector, the control parameters between adjacent control units are iteratively calculated for consistency. Through the consistency algorithm, the parameters between different control units can be globally coordinated and the global cooperative control parameters are generated. Based on the global cooperative control parameters, the Lyapunov energy function of the system is constructed and derived. The construction of the Lyapunov energy function is used to analyze the convergence of the distributed control system. The convergence conditions obtained by derivation are used to verify whether the system can achieve stable operation in a dynamically changing environment. According to the convergence conditions, the control parameters are constrained and bounded to generate a control parameter sequence that meets the temperature balance requirements. The control parameter sequence is substituted into the dynamic balance equation group and numerically integrated to obtain the dynamic balance parameters of each control unit. These parameters describe the specific control strategy required for each fan to maintain system balance during dynamic operation. The dynamic balance parameters are input into the state machine controller. The state machine controller switches the heat dissipation mode according to the current temperature threshold, load level and noise limit conditions. The state machine can dynamically switch between the normal heat dissipation mode, emergency heat dissipation mode, low noise mode and maximum heat dissipation mode, and output the corresponding fan speed control signal. Through this control signal, the operating state of the fan is adjusted in real time to ensure efficient heat dissipation under various working conditions while keeping the energy consumption and noise within an acceptable range.

[0034] The dynamic balance parameters of each control unit are input into the multi-mode state machine controller to build a switching rule framework based on the hierarchical state machine structure. The hierarchical state machine structure can effectively manage the logical relationship between the normal cooling mode, emergency cooling mode, low noise mode and maximum cooling mode, and generate a mode switching condition matrix through rule definition and state transfer control. The matrix defines the transfer conditions between different modes. The system temperature data is processed in a hierarchical manner. The temperature data is divided into four thresholds, including [T1=45℃, T2=60℃, T3=75℃, T4=85℃], and a temperature mode mapping vector is established. Under this division, when the temperature is in the [T1, T2] interval, it is mapped to the normal cooling mode, when it is in the [T2, T3] interval, it corresponds to the emergency cooling mode, and when it is in the [T3, T4] interval, it is mapped to the maximum cooling mode. The hierarchical mapping can ensure that the system switches to the most suitable cooling mode under different temperature conditions and realizes effective temperature control. Based on the temperature mode division, the system load rate is divided into thresholds. The load rate is divided into three levels, including the first load level threshold L1=40%, the second load level threshold L2=70% and the third load level threshold L3=90%. According to these load threshold division rules, when the load rate is lower than L1, the system is set to low noise mode, and when the load rate exceeds L3, it enters the maximum heat dissipation mode. Through the mapping process, a load mode mapping vector is generated. The system noise signal is processed, and the noise limit range is defined by the double threshold limiting method. The noise upper limit of the low noise mode is set to N1=35dB, while the noise upper limit of the conventional heat dissipation mode is set to N2=45dB. The limiting process generates a noise constraint vector, which enables the system to strictly limit the noise level in different modes to meet the requirements of the operating environment. The temperature mode mapping vector, the load mode mapping vector and the noise constraint vector are input into the mode state combiner, and the transition matrix of the four operating modes is generated through state space mapping. The transfer matrix describes the dynamic conversion rules of the system between different modes and clarifies the possibility and conditions of mode switching. In order to optimize the mode switching logic, the transfer matrix is ​​feature analyzed, and the state transition probability matrix between modes is calculated using the maximum likelihood estimation method. The state transition probability matrix provides the switching probability between modes, which is used to guide the system to select the best switching path under different states. Based on the state transition probability matrix, a Viterbi state search tree containing four operating modes is constructed. The Viterbi algorithm can find the optimal mode switching sequence in the search tree through dynamic programming method. This sequence is the best mode switching path for the system under the current operating conditions, which can optimize the system performance on the basis of balancing temperature, load and noise targets. The optimal mode switching sequence is input into the control signal generator. The control signal generator generates the corresponding fan speed control signal for each mode through table lookup and interpolation operations based on the fan speed control strategy of different operating modes.These control signals are dynamically allocated to achieve precise control of fan speed, allowing the system to maintain efficient cooling performance under different operating conditions while minimizing energy consumption and noise levels.

[0035] In the embodiment of the present invention, by constructing a dynamic state space model and an expectation-maximization sliding window iterative extended Kalman filter, the system state is accurately estimated, which effectively solves the problem of inaccurate state estimation in traditional control methods; a three-stage cascade filter design is adopted to achieve hierarchical filtering of system noise, significantly reduce the fan operation noise, and improve the stability of the system; an airflow coupling influence coefficient matrix is ​​introduced, and multi-objective optimization is achieved through Pareto optimization solution, which reduces energy consumption while ensuring heat dissipation efficiency; a distributed collaborative optimization strategy is adopted to solve the control conflict problem in a multi-fan system and achieve dynamic balance among the control units; a control strategy based on a multi-mode state machine is designed, and smooth switching between different operating modes is achieved through a dynamic programming algorithm, thereby avoiding system oscillation during mode switching.

[0036] In a specific embodiment, the process of executing step S1 may specifically include the following steps:

[0037] Data is collected from n temperature sensors of the multi-fan cooling system to obtain a temperature sampling data sequence, and the temperature sampling data sequence is input into a data buffer for sorting and grouping to obtain a temperature observation vector;

[0038] The speed signals of multiple cooling fans in the multi-fan cooling system are sampled at high speed to obtain a fan speed state vector, and the fan speed state vector is input into a state transfer calculation unit, and the state transfer relationship is fitted and calculated by the least square method to obtain a state transfer matrix;

[0039] The vibration and airflow noise of multiple cooling fans are sensed and collected, and the noise signal is separated in the frequency domain to obtain the system noise vector, which is then input into the control matrix calculation unit. The input response characteristics are calculated using the optimal control theory to obtain the control matrix.

[0040] According to the temperature observation vector and the fan speed state vector, the observation matrix is ​​optimized and calculated by the Kalman gain criterion to obtain the observation matrix;

[0041] The state transfer matrix, control matrix, observation matrix and system noise vector are combined into state space equations to obtain the system operation state equation and temperature observation equation.

[0042] Specifically, for the multi-fan cooling system Temperature sensors are used to collect data. Assume that the sensor sampling frequency is , each sensor will generate a temperature sampling point during the sampling period. After that, we get a length of The temperature sampling data sequence is recorded as ,in Indicates The sensor in The collected raw data is input into the data buffer for preprocessing. The sorting operation arranges the data according to the timestamp, and the grouping is clustered according to the sensor location or physical area. , which is the temperature observation vector. At the same time, the speed signals of multiple cooling fans in the system are sampled at high speed. Assume that the number of fans is , the sampled speed signal sequence is recorded as ,in Indicates The fan is in The speed value of each sampling point. These data are organized into a fan speed state vector , and input it into the state transfer calculation unit for analysis. Based on the time series relationship of multiple fan speeds, the state transfer relationship of the system is fitted using the least squares method. Assume that the state transfer model is a linear relationship:

[0043] ;

[0044] in, is the state transfer matrix, which describes the evolution law of the system state; is the process noise, which represents the uncontrollable random disturbance. ,have to:

[0045] ;

[0046] The vibration and airflow noise signals of multiple cooling fans in the system are collected by sensing, and the signals are separated by frequency domain analysis. Assume that the collected noise signal is , through Fourier transform, we can get the spectrum distribution of noise After extracting the features of the noise spectrum, the system noise vector is constructed , which describes the interference characteristics of noise on the system state and observation. The noise vector is input into the control matrix calculation unit, combined with the optimal control theory, and the control matrix is ​​obtained by optimizing the input response characteristics of the system. , satisfying the following relationship:

[0047] ;

[0048] in is the control input vector, which represents the externally applied regulation action. and the fan speed state vector , using the Kalman gain criterion to measure the observation matrix Perform optimization calculation. The observation equation is expressed as:

[0049] ;

[0050] in is the actual observed value, is the observation matrix. By minimizing the observation error covariance, the optimized Ensure that the mapping relationship between the observed value and the true state is optimal. , control matrix , observation matrix and the system noise vector Combine them to complete the construction of the state space model of the system. The operating state equation is:

[0051] ;

[0052] The temperature observation equation is:

[0053] .

[0054] In a specific embodiment, the process of executing step S2 may specifically include the following steps:

[0055] Input the system operation state equation and temperature observation equation into the expectation maximization sliding window iterative extended Kalman filter, set the sliding window length, and determine the maximum number of iterations to obtain the sliding window parameters;

[0056] Based on the sliding window parameters, the system state vector is initialized and calculated to obtain the initial state vector and initial covariance matrix;

[0057] Input the initial state vector and the initial covariance matrix into the E-step calculation unit, calculate the conditional expected value of the state vector, obtain the expected value of the state vector, and perform covariance matrix calculation on the expected value of the state vector to obtain the covariance matrix of the current state;

[0058] Input the expected value and covariance matrix of the state vector into the M-step calculation unit, perform update calculation on the system parameters, and obtain the updated state transfer matrix and the updated observation matrix;

[0059] According to the updated state transfer matrix and the updated observation matrix C, the covariance matrix of the system noise and the observation noise is calculated to obtain the noise covariance matrix;

[0060] The expected value of the state vector and the noise covariance matrix are input into the prediction unit of the extended Kalman filter to calculate the state prediction value at the next moment to obtain the system state estimation value, and based on the system state estimation value, the covariance of the prediction error is calculated to obtain the prediction error covariance matrix.

[0061] Specifically, the system operation state equation and temperature observation equation are input into the expectation maximization sliding window iterative extended Kalman filter, and the length of the sliding window is set. and the maximum number of iterations These parameters determine the calculation range and convergence accuracy of the filter when processing real-time dynamic data. The length of the sliding window Defines the time period that the filter refers to at the current moment, and the maximum number of iterations It is used to limit the number of calculations for each iteration update to ensure that the algorithm is completed within a reasonable time. The system state vector is initialized and calculated in the sliding window parameter setting. Assume that the state equation of the system is ,in is the system state vector, which represents the state at the current moment; is the state transfer matrix, describing the dynamic changes of the state; is the control matrix, representing the external control input Impact on the system; is process noise, representing uncontrollable random disturbance, satisfying zero-mean Gaussian distribution, and covariance is The observation equation is ,in is the observation vector, is the observation matrix, describing the relationship between the observation value and the system state, is the observation noise, which satisfies the zero-mean Gaussian distribution and has a covariance of . In the initialization calculation, it is the system state vector Assign an initial estimate and the initial covariance matrix , which are set by system historical data or prior knowledge. The initial state vector represents the estimated value of the system state at the initial time, and This describes the uncertainty of this estimate. and the initial covariance matrix Input to the E-step calculation unit of the expectation maximization (EM) algorithm. The E-step uses the data in the sliding window to calculate the conditional expectation value of the state vector , that is, to estimate the state vector when the observed value is known. The calculation formula is:

[0062] ;

[0063] in is the predicted state value, is the Kalman gain, defined as:

[0064] ;

[0065] is the prediction error covariance matrix. Through the above formula, we can get the conditional expected value of the state vector and the updated covariance matrix , respectively represent the estimated value of the current state and its uncertainty. The expected value of the state vector and the covariance matrix Enter the M-step calculation unit. In the M-step, the state transfer matrix is ​​calculated by maximizing the log-likelihood function. , control matrix and the observation matrix Perform iterative optimization. It is expressed as:

[0066] ;

[0067] Similarly, the observation matrix and process noise covariance , observation noise covariance It can also be updated by the optimal estimation method. , , , the system noise covariance matrix and the observation noise covariance matrix Calculate and optimize the state estimation performance of the system. The calculation formula of the noise covariance matrix is:

[0068] ;

[0069] ;

[0070] The expected value of the state vector and the noise covariance matrix , Input the prediction unit of the extended Kalman filter to calculate the state prediction value at the next moment. The prediction formula is:

[0071] ;

[0072] And calculate the prediction error covariance matrix based on the predicted values:

[0073] .

[0074] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0075] According to the estimated value of the system state, the power spectrum density analysis of the system noise spectrum is performed to obtain the noise frequency distribution characteristics, and based on the noise frequency distribution characteristics, the cutoff frequency of the low-pass filter is optimized and calculated to obtain the transfer function of the first-stage filter;

[0076] The prediction error covariance matrix is ​​input into the bandpass filter design unit, and the passband range is dynamically adjusted through the frequency band optimization algorithm to obtain the transfer function of the second-stage filter;

[0077] Performing a cascade operation on the transfer function of the first-stage filter and the transfer function of the second-stage filter to obtain a cascade transfer function;

[0078] The cascade transfer function is input into the adaptive filter design unit, and the filter coefficients are iteratively optimized by the minimum mean square error criterion to obtain the transfer function of the third-order filter;

[0079] According to the transfer function of the third-stage filter, the filter response characteristics are analyzed in the frequency domain to obtain the system frequency response curve, and based on the system frequency response curve, the stability of the series structure of the three-stage filter is analyzed to obtain the system transfer function;

[0080] The system transfer function and the system state estimation value are combined to obtain the system multi-order filtering control model.

[0081] Specifically, the power spectrum density analysis of the system noise spectrum is performed according to the system state estimation value, so as to extract the frequency distribution characteristics of the noise and provide a basis for the design of the low-pass filter. Assume that the noise signal is , its power spectrum density is obtained by fast Fourier transform, the formula is:

[0082] ;

[0083] in, is the power spectral density of the noise, is the frequency, is the signal duration. , determine the main frequency distribution range of the noise, such as high frequency noise in Based on this information, the cutoff frequency of the low-pass filter is optimized. , design the transfer function of the first-stage filter The transfer function of a low-pass filter takes the form of a first-order or second-order filter, for example:

[0084] ;

[0085] in, is the optimized cutoff frequency, is an imaginary unit. This transfer function is used to suppress noise above The frequency components of the prediction error covariance matrix are The input bandpass filter design unit dynamically adjusts the passband range of the filter through the frequency band optimization algorithm. The prediction error covariance matrix describes the statistical characteristics of the state estimation error, and its eigenvalue Reflects the error distribution in different frequency bands. Assume that the passband range of the bandpass filter is , then the optimization algorithm is used to determine and The value of makes the bandpass filter transfer function Satisfies maximum signal fidelity and minimum error propagation, the form is:

[0086] ;

[0087] The transfer function of the first-stage filter is The transfer function of the second stage filter is Perform cascade operations to obtain the joint transfer function The concatenation operation is done by simple multiplication:

[0088] ;

[0089] This transfer function integrates the functions of low-pass and band-pass filtering, and achieves dual suppression of high-frequency noise and errors in specific frequency bands. Input to the adaptive filter design unit. The adaptive filter iteratively optimizes the filter coefficients through the minimum mean square error criterion to dynamically adapt to the changes of signals and noise. Assume that the coefficient vector of the adaptive filter is , the input signal is , the output is:

[0090] ;

[0091] The error is:

[0092] ;

[0093] in, is the expected signal. The minimum mean square error update rule is:

[0094] ;

[0095] in, is the step size factor, which is used to control the convergence speed and stability. Through multiple iterations, the optimized adaptive filter transfer function is obtained. . Based on the transfer function of the third-order filter , perform frequency domain analysis on the filter response characteristics and calculate the system frequency response curve ,in:

[0096] ;

[0097] The frequency response curve reveals the filter's amplification or attenuation characteristics for different frequency signals. Based on this curve, the stability of the series structure of the three-stage filter is analyzed. The stability analysis is completed by pole distribution to ensure that all poles are located within the unit circle (discrete system) or the left half plane (continuous system). The system transfer function and the estimated system state Perform combination operations to generate a multi-order filtering control model. Assume that the output after filtering is ,but:

[0098] .

[0099] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0100] Perform spatial decomposition operation on the system transfer function in the system multi-order filter control model to obtain the airflow characteristic function of each fan, and input the airflow characteristic function into the airflow coupling analysis unit to obtain the interaction coefficient between fans;

[0101] According to the interaction coefficients between fans, the fluid dynamics numerical calculations are performed on the airflow velocity and pressure distribution of each fan to obtain the airflow coupling influence coefficient matrix;

[0102] The airflow coupling influence coefficient matrix is ​​input into the energy consumption modeling unit, and the total energy consumption of the fan group is integrated and calculated according to the fan power characteristic curve to obtain the energy consumption target function;

[0103] Based on the airflow coupling influence coefficient matrix, the acoustic characteristics of the aerodynamic noise and mechanical noise of the fan group are analyzed, and the noise objective function is obtained through weighted superposition calculation;

[0104] The airflow coupling influence coefficient matrix is ​​input into the heat dissipation modeling unit, and the heat dissipation and temperature distribution of the system are calculated according to the thermodynamic transfer equation to obtain the heat dissipation efficiency objective function, and the Pareto objective function is constructed based on the energy consumption objective function, the noise objective function and the heat dissipation efficiency objective function;

[0105] The Pareto objective function is input into the gradient optimization unit, and the non-dominated optimal solution sequence is obtained through iterative calculation of different weight combinations. The non-dominated optimal solution sequence is transformed by parameter mapping to obtain the fan control optimization parameter set of fan speed, airflow angle and power distribution.

[0106] Specifically, the system transfer function is extracted from the multi-order filter control model of the system , which describes the response behavior of the entire cooling system to the input signal. is a frequency variable, Contains the system's amplification or attenuation characteristics for signals of different frequencies. In order to describe the independent airflow characteristics of each fan, the system transfer function is spatially decomposed. Assume that the system contains The overall transfer function of a fan is expressed as:

[0107] ;

[0108] in, It is The airflow characteristic function of each fan describes the dynamic response behavior of the fan. By decomposing, the overall response of the system is analyzed into the independent airflow characteristics of each fan. The airflow coupling relationship between fans is determined by the airflow velocity. and pressure During the analysis, the airflow is modeled based on the governing equations of fluid dynamics, such as the Navier-Stokes equations:

[0109] ;

[0110] in, is the fluid density, is the air velocity vector, It's pressure. is the fluid viscosity, is the volume force. Through numerical calculation, the airflow velocity and pressure distribution are obtained, and the airflow coupling influence coefficient matrix describing the airflow coupling relationship between fans is generated. , whose elements Indicates Fan pair The airflow coupling influence coefficient matrix Input energy consumption modeling unit, combined with the power characteristic curve of the fan Calculate the total energy consumption. Assuming the fan Power consumption and air flow velocity The total energy consumption of the fan group is calculated by integration:

[0111] ;

[0112] in, is the running time, Indicates fan The power characteristic curve is shown in Figure 2. In the energy consumption calculation, the airflow coupling influence coefficient matrix is ​​analyzed to analyze the aerodynamic noise and mechanical noise characteristics of the fan group. and pressure The noise power spectrum density of each fan is calculated by combining the acoustic model (such as the fluid dynamics acoustic equation) The noise objective function of the system is obtained by weighted superposition:

[0113] ;

[0114] in, It's a fan The noise weight, is the frequency range. The airflow coupling influence coefficient matrix is ​​input into the heat dissipation modeling unit, and the heat dissipation efficiency of the system is calculated based on the thermodynamic transfer equation. Assume that the heat dissipation of the system With fan air speed and temperature difference There is a linear relationship, so the heat dissipation efficiency objective function is expressed as:

[0115] ;

[0116] Combined energy consumption objective function , Noise objective function And the heat dissipation efficiency objective function , construct the Pareto objective function:

[0117] ;

[0118] in, , , is the weight coefficient used to balance the priorities of energy consumption, noise and heat dissipation efficiency. Input the gradient optimization unit, and through iterative calculations of different weight combinations, a multi-objective optimization algorithm (such as NSGA-II or weighted gradient descent method) is used to obtain a sequence of non-dominated optimal solutions. For each non-dominated optimal solution, the fan speed is determined through parameter mapping transformation. , airflow angle and power distribution , and finally generate the fan control optimization parameter set.

[0119] In a specific embodiment, executing the cooling fan multi-order filtering control method further includes the following steps:

[0120] The fan control optimization parameter set is input into the distributed computing unit, the control parameters are decoupled to obtain the initial control parameter sets of multiple control units, and the temperature sensor signals of multiple control units are collected and neighborhood information is exchanged to obtain the temperature state information interaction matrix;

[0121] Based on the temperature state information interaction matrix, the temperature coupling strength between control units is quantitatively analyzed, and a distributed communication topology is constructed through a weighted network to obtain the temperature coupling constraint matrix between control units.

[0122] The temperature coupling constraint matrix is ​​input into the local controller, and the fan speed and airflow direction of each control unit are subjected to constraint optimization calculation to obtain the local control optimization vector;

[0123] According to the local control optimization vector, the control parameters between adjacent control units are iteratively calculated to obtain the global coordinated control parameters;

[0124] Based on the global cooperative control parameters, the Lyapunov energy function of the system is constructed and derived to obtain the convergence conditions of the distributed control system.

[0125] According to the convergence conditions, the control parameters are constrained and processed to obtain a control parameter sequence that satisfies the temperature balance. The control parameter sequence is substituted into the dynamic balance equation group for numerical integration and solution to obtain the dynamic balance parameters of each control unit.

[0126] The dynamic balance parameters of each control unit are input into the state machine controller, and the normal cooling mode, emergency cooling mode, low noise mode and maximum cooling mode are switched and judged according to the temperature threshold, load level and noise limit to obtain the fan speed control signal.

[0127] Specifically, the fan control optimization parameter set is recorded as , where each Fan included Speed , airflow angle and power distribution These parameters are input into the distributed computing unit, and the parameters of each control unit are separated through decoupling processing to generate the initial control parameter set. Decoupling provides each control unit with an independent initial control input, ensuring that the control requirements of each fan can be processed independently in subsequent calculations. The temperature sensor signal of each control unit is collected in real time and recorded as ,in It is the control unit No. The collected values ​​of the temperature sensors are collected. At the same time, the temperature data of the adjacent control units are obtained through the neighborhood information exchange mechanism. By aggregating these data, the temperature state information interaction matrix is ​​generated. , where the elements Indicates the control unit and The temperature interaction information strength is based on the temperature state information interaction matrix. , by calculating the temperature coupling strength, the heat conduction characteristics between control units are analyzed. The coupling strength is quantified by the following formula:

[0128] ;

[0129] in, Indicates the control unit right The temperature affects the strength, and The units are and These intensity parameters are constructed as a weighted network to form a distributed communication topology, whose adjacency matrix is ​​the temperature coupling constraint matrix. . The temperature coupling constraint matrix Input the local controller to perform constrained optimization calculations on the fan speed and airflow direction of each control unit. The goal is to minimize local energy consumption while meeting the temperature balance requirements. The local optimization objective function is:

[0130] ;

[0131] in, Yes Unit Energy consumption, is the current temperature, is the target temperature, and is the weight factor. The local control optimization vector is obtained by optimization According to the local control optimization vector, the control parameters between adjacent control units are calculated iteratively to ensure global coordination. The update formula of the consistency iteration is:

[0132] ;

[0133] in, Indicates the control unit Neighborhood of is the step size parameter. Through multiple iterations, the global collaborative control parameters are generated. Based on the global cooperative control parameters, the Lyapunov energy function of the system is constructed , to analyze the convergence of distributed control systems. The Lyapunov energy function is defined as:

[0134] ;

[0135] The convergence condition is obtained by derivation:

[0136] ;

[0137] in, is the convergence rate. When this condition is met, the system can achieve global stability. According to the convergence condition, the control parameters are constrained and processed to generate a control parameter sequence that satisfies temperature equilibrium. . Substituting these parameters into the dynamic equilibrium equations:

[0138] ;

[0139] The dynamic balance parameters of each control unit are obtained through numerical integration. The dynamic balance parameters are input into the state machine controller, and the normal cooling mode, emergency cooling mode, low noise mode and maximum cooling mode are switched according to the temperature threshold, load level and noise limit to obtain the fan speed control signal.

[0140] In a specific embodiment, the execution step inputs the dynamic balance parameters of each control unit into the state machine controller, and switches and judges the normal cooling mode, the emergency cooling mode, the low noise mode and the maximum cooling mode according to the temperature threshold, the load level and the noise limit, and the process of obtaining the fan speed control signal can specifically include the following steps:

[0141] The dynamic balance parameters of each control unit are input into the multi-mode state machine controller, and the switching rules of the conventional cooling mode, the emergency cooling mode, the low noise mode and the maximum cooling mode are constructed through the hierarchical state machine structure to obtain the mode switching condition matrix;

[0142] The system temperature data is divided into four threshold levels [T1=45℃, T2=60℃, T3=75℃, T4=85℃], and the normal cooling mode is set according to the temperature range [T1, T2], the emergency cooling mode is set for [T2, T3], and the maximum cooling mode is set for [T3, T4], and the temperature mode mapping vector is obtained;

[0143] Based on the temperature mode mapping vector, the system load rate is divided into three threshold levels [L1=40%, L2=70%, L3=90%], L1 is the first load level threshold, L2 is the second load level threshold, L3 is the third load level threshold, and the low noise mode is set below the first load level threshold L1, and the maximum heat dissipation mode is set above the third load level threshold L3, to obtain the load mode mapping vector;

[0144] According to the load mode mapping vector, the noise signal is limited by double thresholds [N1=35dB, N2=45dB], and N1 is the noise upper limit of the low noise mode, and N2 is the noise upper limit of the conventional cooling mode, and the noise constraint vector is obtained;

[0145] The temperature mode mapping vector, the load mode mapping vector and the noise constraint vector are input into the mode state combiner, and the transfer matrices of the four operation modes are obtained through state space mapping;

[0146] The characteristic analysis of the transfer matrix of the four operating modes is carried out, and the state transition probability matrix between the modes is obtained by maximum likelihood estimation. Based on the state transition probability matrix, a Viterbi state search tree containing the four operating modes is constructed, and the optimal mode switching sequence is obtained by a dynamic programming algorithm.

[0147] The optimal mode switching sequence is input into the control signal generator, and the fan speed control signal is obtained by table lookup and interpolation operation according to the fan speed control strategy of different operation modes.

[0148] Specifically, the dynamic balance parameters of each control unit Input to the multi-mode state machine controller, where Indicates the control unit The multi-mode state machine defines the switching rules of different modes through a hierarchical structure and describes the conditions for mode switching in matrix form. The system temperature data is processed and the temperature distribution of the system is assumed to be ,in For the The current temperature of each control unit. The temperature data is divided into four threshold ranges С. С. С and C. Based on these thresholds, construct the temperature pattern mapping vector , defined as follows:

[0149] ;

[0150] The mapping ensures that the system can dynamically select the appropriate cooling mode according to the current temperature, thereby achieving effective temperature control. After completing the temperature mode mapping, analyze the system load rate. Assume that the system load rate is ,in Indicates The current load rate of each control unit. The load rate is divided into three thresholds: , , Based on these thresholds, a load pattern mapping vector is constructed :

[0151] ;

[0152] This mapping ensures that the system prioritizes low noise mode under low load conditions and prioritizes maximum cooling mode under high load conditions. , limit the noise signal. Assume the noise signal of the system is , limiting the noise to the double threshold range dB and dB. Get the noise constraint vector :

[0153] ;

[0154] Map the temperature mode to a vector , load pattern mapping vector and the noise constraint vector Input mode state combiner. Generate mode transfer matrix through state space mapping ,in Representation Mode To Mode The mode transfer matrix is ​​used to describe the dynamic switching rules of the four operating modes. The mode transfer matrix is ​​analyzed and the maximum likelihood estimation method is used to calculate the state transition probability matrix between modes. ,in Indicates slave mode Transfer to mode The probability of , build a Viterbi state search tree containing four operating modes. Use the dynamic programming algorithm to find the optimal mode switching sequence in the search tree ,in Indicates at time The optimal mode under the optimal mode. The input control signal generator generates the fan speed control signal through table lookup and interpolation operation according to the fan speed control strategy of different operation modes. The interpolation calculation realizes the smooth transition of the speed and ensures the stability and efficiency of the system operation.

[0155] The above describes the multi-order filtering control method of the cooling fan in the embodiment of the present invention. The following describes the multi-order filtering control device of the cooling fan in the embodiment of the present invention. Figure 2 , an embodiment of the multi-stage filtering control device for a cooling fan in an embodiment of the present invention includes:

[0156] The acquisition module is used to collect the operating parameters of the multi-fan cooling system to build a dynamic state space model and obtain the system operation state equation and temperature observation equation;

[0157] A calculation module is used to input the system operation state equation and the temperature observation equation into the expectation maximization sliding window iterative extended Kalman filter, and obtain the system state estimation value and the prediction error covariance matrix by executing the conditional expectation calculation of the E step and the parameter iterative update of the M step;

[0158] A design module is used to perform a three-stage series design of the cutoff frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter according to the system state estimation value and the prediction error covariance matrix, so as to obtain a multi-order filtering control model of the system;

[0159] The solution module is used to calculate the airflow coupling influence coefficient matrix between multiple fans based on the system multi-order filtering control model, and optimize the total energy consumption, noise level and heat dissipation efficiency through the gradient iteration method to obtain the fan control optimization parameter set.

[0160] Through the coordinated cooperation of the above-mentioned components, by constructing a dynamic state space model and an expectation-maximization sliding window iterative extended Kalman filter, the system state is accurately estimated, effectively solving the problem of inaccurate state estimation in traditional control methods; a three-stage cascade filter design is adopted to realize hierarchical filtering of system noise, significantly reduce the fan operation noise, and improve the stability of the system; the airflow coupling influence coefficient matrix is ​​introduced, and multi-objective optimization is realized through Pareto optimization solution, which reduces energy consumption while ensuring heat dissipation efficiency; a distributed collaborative optimization strategy is adopted to solve the control conflict problem in the multi-fan system and achieve dynamic balance among the control units; a control strategy based on a multi-mode state machine is designed, and smooth switching between different operating modes is realized through a dynamic programming algorithm, avoiding system oscillation during mode switching.

[0161] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0162] Those skilled in the art will understand that Figure 3The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0163] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0164] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0165] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0166] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0167] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-order filtering control method for a cooling fan, characterized in that: The method comprises: The operating parameters of the multi-fan cooling system are collected to construct a dynamic state space model, and the system operating state equation and temperature observation equation are obtained; Inputting the system operation state equation and the temperature observation equation into an expectation maximization sliding window iterative extended Kalman filter, and obtaining a system state estimation value and a prediction error covariance matrix by performing conditional expectation calculation and parameter iterative update; According to the system state estimation value and the prediction error covariance matrix, the cutoff frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter are designed in three stages in series to obtain a system multi-order filtering control model; specifically including: according to the system state estimation value, a power spectral density analysis is performed on the system noise spectrum to obtain the noise frequency distribution characteristics, and based on the noise frequency distribution characteristics, the cutoff frequency of the low-pass filter is optimized and calculated to obtain the transfer function of the first-stage filter; the prediction error covariance matrix is ​​input into the band-pass filter design unit, and the passband range is dynamically adjusted through the frequency band optimization algorithm to obtain the transfer function of the second-stage filter. ; Performing a cascade operation on the transfer function of the first-stage filter and the transfer function of the second-stage filter to obtain a cascade transfer function; inputting the cascade transfer function into an adaptive filter design unit, iteratively optimizing the filter coefficients by a minimum mean square error criterion, and obtaining a transfer function of a third-stage filter; performing a frequency domain analysis on the filter response characteristics according to the transfer function of the third-stage filter to obtain a system frequency response curve, and based on the system frequency response curve, performing a stability analysis on the series structure of the three-stage filter to obtain a system transfer function; performing a combined operation on the system transfer function and the system state estimation value to obtain a system multi-order filtering control model; Based on the multi-order filtering control model of the system, the airflow coupling influence coefficient matrix between multiple fans is calculated, and the total energy consumption, noise level and heat dissipation efficiency are optimized through the gradient iteration method to obtain the fan control optimization parameter set; specifically including: performing spatial decomposition operation on the system transfer function in the multi-order filtering control model of the system to obtain the airflow characteristic function of each fan, and inputting the airflow characteristic function into the airflow coupling analysis unit to obtain the interaction coefficient between the fans; according to the interaction coefficient between the fans, the airflow velocity and pressure distribution of each fan are numerically calculated by fluid dynamics to obtain the airflow coupling influence coefficient matrix; the airflow coupling influence coefficient matrix is ​​input into the energy consumption modeling unit, and the total energy consumption of the fan group is integrated and calculated according to the fan power characteristic curve. , and obtain the energy consumption objective function; based on the airflow coupling influence coefficient matrix, the acoustic characteristics of the aerodynamic noise and mechanical noise of the fan group are analyzed, and the noise objective function is obtained through weighted superposition calculation; the airflow coupling influence coefficient matrix is ​​input into the heat dissipation modeling unit, and the heat dissipation and temperature distribution of the system are calculated according to the thermodynamic transfer equation to obtain the heat dissipation efficiency objective function, and a Pareto objective function is constructed based on the energy consumption objective function, the noise objective function and the heat dissipation efficiency objective function; the Pareto objective function is input into the gradient optimization unit, and a non-dominated optimal solution sequence is obtained through iterative calculation of different weight combinations, and the non-dominated optimal solution sequence is transformed by parameter mapping to obtain a fan control optimization parameter set of fan speed, airflow angle and power distribution.

2. The multi-order filtering control method for a cooling fan according to claim 1, characterized in that: The operating parameters of the multi-fan cooling system are collected to construct a dynamic state space model to obtain the system operating state equation and temperature observation equation, including: Collecting data from n temperature sensors of the multi-fan cooling system to obtain a temperature sampling data sequence, and inputting the temperature sampling data sequence into a data buffer for sorting and grouping to obtain a temperature observation vector; The speed signals of the multiple cooling fans in the multi-fan cooling system are sampled at high speed to obtain a fan speed state vector, and the fan speed state vector is input into a state transfer calculation unit, and the state transfer relationship is fitted and calculated by the least square method to obtain a state transfer matrix; The vibration and airflow noise of the plurality of cooling fans are sensed and collected, and the noise signal is separated in the frequency domain to obtain a system noise vector, and the system noise vector is input into a control matrix calculation unit, and the input response characteristic is calculated by optimal control theory to obtain a control matrix; According to the temperature observation vector and the fan speed state vector, an observation matrix is ​​optimized and calculated by using a Kalman gain criterion to obtain an observation matrix; The state transfer matrix, the control matrix, the observation matrix and the system noise vector are subjected to state space equation combination operation to obtain a system operation state equation and a temperature observation equation.

3. The multi-order filtering control method for a cooling fan according to claim 2, characterized in that: The system operation state equation and the temperature observation equation are input into the expectation maximization sliding window iterative extended Kalman filter, and the system state estimation value and the prediction error covariance matrix are obtained by performing conditional expectation calculation and parameter iterative update, including: Input the system operation state equation and the temperature observation equation into the expectation maximization sliding window iterative extended Kalman filter, set the sliding window length, and determine the maximum number of iterations to obtain the sliding window parameters; Based on the sliding window parameters, the system state vector is initialized and calculated to obtain an initial state vector and an initial covariance matrix; Inputting the initial state vector and the initial covariance matrix into the E-step calculation unit, calculating the conditional expected value of the state vector to obtain the expected value of the state vector, and performing covariance matrix calculation on the expected value of the state vector to obtain the covariance matrix of the current state; Input the expected value of the state vector and the covariance matrix into the M-step calculation unit, perform update calculation on the system parameters, and obtain an updated state transfer matrix and an updated observation matrix; Calculate the covariance matrix of system noise and observation noise according to the updated state transfer matrix and the updated observation matrix C to obtain a noise covariance matrix; The expected value of the state vector and the noise covariance matrix are input into the prediction unit of the extended Kalman filter, the state prediction value at the next moment is calculated to obtain the system state estimation value, and based on the system state estimation value, the covariance of the prediction error is calculated to obtain the prediction error covariance matrix.

4. The multi-order filtering control method for a cooling fan according to claim 1, characterized in that: The multi-order filtering control method for the cooling fan further includes: Input the fan control optimization parameter set into a distributed computing unit, perform decoupling processing on the control parameters to obtain initial control parameter sets of multiple control units, and perform data collection and neighborhood information exchange on temperature sensor signals of the multiple control units to obtain a temperature state information interaction matrix; Based on the temperature state information interaction matrix, the temperature coupling strength between the control units is quantitatively analyzed, and a distributed communication topology is constructed through a weighted network to obtain a temperature coupling constraint matrix between the control units; The temperature coupling constraint matrix is ​​input into a local controller, and constraint optimization calculation is performed on the fan speed and airflow direction of each control unit to obtain a local control optimization vector; According to the local control optimization vector, the control parameters between adjacent control units are iteratively calculated to obtain global coordinated control parameters; Based on the global collaborative control parameters, the Lyapunov energy function of the system is constructed and derived to obtain the convergence condition of the distributed control system; According to the convergence conditions, the control parameters are constrained and processed to obtain a control parameter sequence that satisfies the temperature balance, and the control parameter sequence is substituted into the dynamic balance equation group for numerical integration and solution to obtain the dynamic balance parameters of each control unit; The dynamic balance parameters of each control unit are input into the state machine controller, and the normal cooling mode, emergency cooling mode, low noise mode and maximum cooling mode are switched and judged according to the temperature threshold, load level and noise limit to obtain the fan speed control signal.

5. The multi-order filtering control method for a cooling fan according to claim 4, characterized in that: The dynamic balance parameters of each control unit are input into the state machine controller, and the conventional cooling mode, the emergency cooling mode, the low noise mode and the maximum cooling mode are switched and judged according to the temperature threshold, the load level and the noise limit to obtain the fan speed control signal, including: The dynamic balance parameters of each control unit are input into a multi-mode state machine controller, and the switching rules of the conventional heat dissipation mode, the emergency heat dissipation mode, the low noise mode and the maximum heat dissipation mode are constructed through a hierarchical state machine structure to obtain a mode switching condition matrix; The system temperature data is divided into four threshold levels [T1=45℃, T2=60℃, T3=75℃, T4=85℃], and the normal cooling mode is set according to the temperature range [T1, T2], the emergency cooling mode is set for [T2, T3], and the maximum cooling mode is set for [T3, T4], and the temperature mode mapping vector is obtained; Based on the temperature mode mapping vector, the system load rate is divided into three threshold levels [L1=40%, L2=70%, L3=90%], L1 is the first load level threshold, L2 is the second load level threshold, L3 is the third load level threshold, and a low noise mode is set below the first load level threshold L1, and a maximum heat dissipation mode is set above the third load level threshold L3, to obtain a load mode mapping vector; According to the load mode mapping vector, the noise signal is limited by double thresholds [N1=35dB, N2=45dB], and N1 is the noise upper limit of the low noise mode, and N2 is the noise upper limit of the conventional heat dissipation mode, to obtain a noise constraint vector; Input the temperature mode mapping vector, the load mode mapping vector and the noise constraint vector into a mode state combiner, and obtain transfer matrices of four operation modes through state space mapping; Performing feature analysis on the transfer matrices of the four operating modes, obtaining the state transition probability matrix between the modes by maximum likelihood estimation, and constructing a Viterbi state search tree containing the four operating modes based on the state transition probability matrix, and obtaining the optimal mode switching sequence by a dynamic programming algorithm; The optimal mode switching sequence is input into a control signal generator, and a fan speed control signal is obtained by table lookup and interpolation operation according to fan speed control strategies of different operation modes.

6. A multi-stage filter control device for a cooling fan, characterized in that: Used to execute the multi-order filtering control method for a cooling fan according to any one of claims 1 to 5, the cooling fan multi-order filtering control device comprises: The acquisition module is used to collect the operating parameters of the multi-fan cooling system to build a dynamic state space model and obtain the system operation state equation and temperature observation equation; A calculation module, used for inputting the system operation state equation and the temperature observation equation into an expectation maximization sliding window iterative extended Kalman filter, and obtaining a system state estimation value and a prediction error covariance matrix by performing conditional expectation calculation and parameter iterative update; A design module is used to perform a three-stage series design on the cutoff frequency of the low-pass filter, the passband range of the band-pass filter and the coefficient of the adaptive filter according to the system state estimation value and the prediction error covariance matrix, so as to obtain a system multi-order filtering control model; The solution module is used to calculate the airflow coupling influence coefficient matrix between multiple fans based on the multi-order filtering control model of the system, and optimize the total energy consumption, noise level and heat dissipation efficiency through the gradient iteration method to obtain the fan control optimization parameter set.

7. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the multi-order filtering control method for the cooling fan described in any one of claims 1 to 5 is implemented. 8 . A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the multi-order filtering control method for a cooling fan according to any one of claims 1 to 5.

Images