Dynamic control method and system for modularized energy-saving array of server fan

Abstract

The invention discloses a dynamic control method and system for a modularized energy-saving array of a server fan, relates to the technical field of server heat dissipation control, and aims to improve the efficiency of the modularized energy-saving array of the server fan through collaborative optimization of multi-dimensional data acquisition and federated learning in combination with a dynamic speed regulation and redundancy control mechanism. The method achieves the precise prediction of the load change, optimizes the distribution of heat dissipation resources, reduces the energy consumption, and improves the heat dissipation efficiency and equipment operation stability.

Description

Technical Field

[0001] This application relates to the field of server heat dissipation control technology, and in particular to a dynamic control method and system for modular energy-saving arrays of server fans. Background Technology

[0002] In today's server operating environment, with the increasing demand for data processing, the power consumption and heat dissipation issues of servers are becoming increasingly prominent. Traditional server fan control methods often use fixed speeds or simple temperature feedback control. This approach makes it difficult to dynamically adjust the fan speed according to the actual load of the server, resulting in the fan still running at a high speed under low load, causing energy waste; while under high load, insufficient heat dissipation may lead to server performance degradation or even failure.

[0003] Furthermore, most existing fan control solutions lack comprehensive analysis of multi-dimensional data such as vibration and airflow, making it difficult to accurately assess the server's cooling needs and fan operating status. At the same time, when dealing with multi-server clusters, the lack of an effective collaborative control mechanism leads to inconsistent fan operating states across servers, further impacting overall cooling performance and energy efficiency.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a dynamic control method and system for a modular energy-saving array of server fans, which has the advantages of achieving precise speed adjustment to reduce energy consumption, improve heat dissipation efficiency and operational stability.

[0006] To achieve the above objectives, this application proposes a dynamic control method for a modular energy-saving array of server fans, the method comprising: Raw sensor data, including temperature data, vibration data, and airflow data, is collected through a sensor array in a modular energy-saving array. The raw sensor data is preprocessed to obtain standardized sensor data. The standardized sensor data is subjected to load prediction processing to obtain predicted load data; Based on the predicted load data, the preset regional temperature difference data, and the historical power consumption data, multi-objective dynamic speed regulation processing is performed to generate fan control command data. Through a federated learning collaboration mechanism, local model parameter data from multiple edge nodes are aggregated on a cloud server to generate global optimization strategy data. The final control signal is generated based on the global optimization strategy data and fan control command data to control the operating status of the fan modular array.

[0007] In one embodiment, the step of preprocessing the raw sensor data to obtain standardized sensor data includes: The vibration data in the original sensor data is subjected to wavelet transform noise reduction processing to obtain noise-reduced vibration data. The temperature data and airflow data in the original sensor data are processed by Kalman filtering to obtain filtered temperature data and filtered airflow data. The noise-reduced vibration data, filtered temperature data, and filtered airflow data are normalized to obtain standardized sensor data.

[0008] In one embodiment, the step of performing load prediction processing on the standardized sensor data to obtain predicted load data includes: Extract historical load characteristic data from the standardized sensor data; The historical load feature data is input into a pre-trained long short-term memory prediction model for time-series prediction processing, and the predicted load data is output.

[0009] In one embodiment, the step of performing multi-objective dynamic speed regulation processing to generate fan control command data based on the predicted load data, preset regional temperature difference data, and historical power consumption data includes: Calculate the current load rate data based on the predicted load data; The current load rate data, the preset regional temperature difference data, and the historical power consumption data are input into the multi-agent collaborative optimization model. The optimization calculation is performed through the multi-agent collaborative optimization model, and the objective function value data is output. Fan control command data is generated based on the objective function value data.

[0010] In one embodiment, the step of performing optimization calculations through the multi-agent collaborative optimization model and outputting objective function value data includes: The fan control agent generates fan action data based on the temperature and vibration data from the standardized sensor data. The DVFS control agent generates voltage-frequency combined action data based on load data and core temperature data; The fan action data and voltage-frequency combined action data are executed together.

[0011] In one embodiment, the step of jointly executing the fan action data and the voltage-frequency combined action data includes: After executing the fan action data and voltage-frequency combined action data, real-time power consumption data is collected; Collect noise sensor data and convert it into linearized noise index data; The standard deviation of the vibration data is calculated to obtain the stability index data. The objective function value is calculated by combining the real-time power consumption data, linearized noise index data, and stability index data according to weights.

[0012] In one embodiment, after the step of acquiring raw sensor data via a sensor array in a modular energy-efficient array, the method further includes: The vibration data in the original sensor data is subjected to spectral analysis to obtain vibration spectral characteristic data; When the vibration spectrum characteristic data exceeds the preset fault threshold, a faulty fan unit identification signal is generated; The hot-swap redundancy control mechanism is triggered based on the faulty fan unit identification signal to activate the backup fan unit.

[0013] In one embodiment, the step of aggregating local model parameter data from multiple edge nodes on a cloud server to generate global optimization strategy data through a federated learning collaborative mechanism includes: Local policy models are trained at each edge node based on standardized sensor data to generate local model parameter data. Differential privacy noise is added to the local model parameter data to generate noise-added parameter data; The noise-added parameter data is homomorphically encrypted to generate encrypted model parameter data. The system aggregates parameter data from multiple encryption models and performs policy update processing to generate global optimization policy data.

[0014] In one embodiment, the method further includes: Collect external ambient temperature data; When the ambient temperature is below the low temperature threshold, the fan is stopped and the phase change material is activated to dissipate heat. When the external ambient temperature exceeds the high temperature threshold, the redundant fan unit data is activated and the liquid cooling auxiliary system is started.

[0015] Furthermore, to achieve the above objectives, this application also proposes a dynamic control system for a modular energy-saving array of server fans. The dynamic control system for a modular energy-saving array of server fans includes: a memory, a processor, and a dynamic control program for a modular energy-saving array of server fans stored in the memory and executable on the processor. The dynamic control program for a modular energy-saving array of server fans is configured to implement the steps of the dynamic control method for a modular energy-saving array of server fans.

[0016] The server fan modular energy-saving array dynamic control method and system proposed in this application achieves accurate prediction of load changes and optimization of heat dissipation resource allocation through multi-dimensional data acquisition and federated learning collaborative optimization, combined with dynamic speed regulation and redundancy control mechanisms. It has the advantages of reducing energy consumption, improving heat dissipation efficiency and equipment operation stability. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the dynamic control method for a modular energy-saving array of server fans in this application. Figure 2 For this application Figure 1 A detailed flowchart of step S200; Figure 3 For this application Figure 1 Detailed flowchart of step S400; Figure 4 For this application Figure 1 Detailed flowchart of step S500; Figure 5 This is a schematic diagram of a structural design for an embodiment of the modular energy-saving array dynamic control system for server fans in this application.

[0020] Explanation of icon numbers: 10. Memory; 20. Processor.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] In existing technologies, server heat dissipation control has long relied on fixed rotation speeds or single temperature feedback mechanisms, which cannot adapt to dynamic load changes. Traditional methods lead to energy waste by continuously operating at high speeds under low loads, while insufficient heat dissipation under high loads may cause equipment failures. Furthermore, the lack of comprehensive analysis of multi-dimensional data such as vibration and airflow makes it difficult to accurately assess heat dissipation requirements. In multi-server cluster scenarios, independent control of each node results in low heat dissipation efficiency and increased overall energy consumption.

[0025] To address the aforementioned issues, the inability of existing technologies to dynamically adjust fan speed prompted the inventors to consider how to combine multi-dimensional data to achieve accurate predictions. Considering the impact of load fluctuations on heat dissipation requirements, introducing a time-series prediction model to anticipate future loads became crucial. The need for multi-objective optimization drove the inventors to explore collaborative speed control mechanisms to balance temperature, power consumption, and stability. The cross-node collaboration challenge was solved using a federated learning framework, achieving global policy optimization.

[0026] Based on this, this application provides a dynamic control method for a modular energy-saving array of server fans, referring to... Figure 1 The dynamic control method for the modular energy-saving array of server fans includes steps S100 to S600, wherein: Step S100: Collect raw sensor data through the sensor group in the modular energy-saving array. The raw sensor data includes temperature data, vibration data, and airflow data. Step S200: Perform data preprocessing on the raw sensor data to obtain standardized sensor data; Step S300: Perform load prediction processing on the standardized sensor data to obtain predicted load data; Step S400: Based on the predicted load data, preset regional temperature difference data and historical power consumption data, perform multi-objective dynamic speed regulation processing to generate fan control command data; Step S500: Through the federated learning collaboration mechanism, the local model parameter data of multiple edge nodes are aggregated on the cloud server to generate global optimization strategy data; Step S600: Generate a final control signal based on the global optimization strategy data and fan control command data to control the operating status of the fan modular array.

[0027] In this embodiment, the sensor group refers to temperature, vibration, and airflow detection devices distributed at different locations on the server. Specifically, it can be implemented using a combination of thermocouples, accelerometers, and flow meters for real-time monitoring of the equipment's operating environment. Data preprocessing includes noise reduction, filtering, and normalization operations. Specifically, wavelet transform can be used to process vibration signals, and Kalman filtering can be used to process temperature and airflow fluctuations, eliminating measurement noise interference. Load prediction processing refers to predicting future loads by analyzing historical data trends. Specifically, a long short-term memory neural network model can be used to process time-series data and capture load change patterns. Multi-objective dynamic speed regulation refers to a speed regulation strategy that simultaneously optimizes temperature control, energy saving, and equipment stability. Specifically, a multi-agent reinforcement learning model can be used to calculate the optimal speed combination. The federated learning collaborative mechanism refers to model training while protecting data privacy. Specifically, homomorphic encryption technology can be used to aggregate local model parameters from each node to form a global optimization strategy.

[0028] In this embodiment, temperature, vibration, and airflow data are collected by sensors. Wavelet transform is used to eliminate mechanical vibration noise, Kalman filtering is employed to smooth temperature and airflow fluctuations, and normalization is used to form standardized data with uniform dimensions. A Long Short-Term Memory (LSTM) model analyzes the load change trend in the standardized data and outputs predicted load values ​​for future periods. A multi-agent optimization model integrates predicted load, regional temperature difference, and power consumption records to calculate the fan speed combination that meets heat dissipation requirements and has the lowest energy consumption. The cloud server collects encrypted local model parameters from each edge node using a federated learning framework, decrypts and aggregates them, and updates the global control strategy. Finally, the control signal integrates local speed control commands with the global optimization strategy to dynamically adjust the start / stop status and speed parameters of each fan module.

[0029] In this embodiment, the solution achieves proactive control through multi-source data fusion and load prediction, and employs a federated learning framework to optimize cluster-level heat dissipation while ensuring data security. Furthermore, it enhances state perception accuracy through multi-sensor data fusion. Thus, this application effectively reduces redundant energy consumption under low load conditions and avoids the risk of insufficient heat dissipation under high load conditions. Multi-dimensional data fusion improves the ability to perceive the device's operating status, making heat dissipation control more precise and reliable. The federated learning mechanism achieves cross-node collaborative optimization while protecting data privacy, significantly improving the overall energy efficiency of large-scale server clusters. The dynamic speed adjustment strategy balances heat dissipation efficiency and energy consumption indicators, extending the lifespan of fan equipment.

[0030] In one feasible implementation, refer to Figure 2 Step S200 includes steps S210 to S230, wherein: Step S210: Perform wavelet transform noise reduction processing on the vibration data in the original sensor data to obtain noise-reduced vibration data; Step S220: Perform Kalman filtering on the temperature data and airflow data in the original sensor data to obtain filtered temperature data and filtered airflow data; Step S230: Normalize the noise-reduced vibration data, filtered temperature data, and filtered airflow data to obtain standardized sensor data.

[0031] In this embodiment, wavelet transform denoising refers to the process of decomposing the vibration signal using wavelet basis functions and removing high-frequency noise components. Specifically, it can be implemented using discrete wavelet transform combined with a threshold shrinkage algorithm. For example, after decomposing the vibration signal into five levels using the Daubechies wavelet basis, soft thresholding is applied to the detail coefficients. This process can effectively eliminate high-frequency interference noise in mechanical vibration signals. Kalman filtering refers to the process of recursively optimizing the observation data of a dynamic system using a state-space model. Specifically, a linear Kalman filter can be used to filter the time-series data of temperature and airflow sensors. For example, state transition equations and observation equations for temperature measurement can be established for prediction and correction. This process can eliminate random measurement errors in temperature and airflow data. Normalization refers to the process of mapping sensor data of different dimensions to a unified numerical range. Specifically, a minimum-maximum normalization method can be used to linearly transform vibration, temperature, and airflow data to the [0,1] interval. For example, scaling can be performed by statistically analyzing the maximum and minimum values ​​of each sensor based on historical data. This process can eliminate dimensional differences between multi-source sensor data.

[0032] In this embodiment, the vibration data is first decomposed into approximation coefficients and detail coefficients using wavelet transform. After removing high-frequency noise components by setting a threshold, the denoised vibration data is reconstructed. Temperature and airflow data are predicted and corrected using Kalman filters, respectively. The current measurement value is optimized using the state estimate from the previous moment to eliminate sensor drift error. The three sets of data after denoising and filtering are then normalized to make the vibration, temperature, and airflow data comparable, forming standardized sensor data as input for subsequent load prediction.

[0033] In this embodiment, the nonlinear characteristics of the vibration signal and the dynamic characteristics of temperature and airflow are matched by a combination of wavelet transform and Kalman filtering, which significantly improves the signal-to-noise ratio of the multi-dimensional sensor data. Thus, the above solution solves the problems of high-frequency noise, dynamic measurement error and inconsistency of dimensions in the original sensor data. The data quality is improved by targeted noise reduction and filtering, and the normalization process ensures that multi-source data participate in the calculation under a unified scale, providing high-precision standardized input data for the subsequent load prediction model.

[0034] In one feasible implementation, the step of performing load prediction processing on the standardized sensor data to obtain predicted load data includes: extracting historical load feature data from the standardized sensor data; inputting the historical load feature data into a pre-trained long short-term memory prediction model for time-series prediction processing; and outputting predicted load data.

[0035] In this embodiment, historical load characteristic data refers to a set of time-series features related to server load status selected from standardized sensor data. Specifically, it can be implemented using a sliding window technique to extract temperature change rate, vibration amplitude trend, and airflow fluctuation pattern over several past time periods. These features can reflect the dynamic changes in server load. The Long Short-Term Memory (LSTM) prediction model refers to a recurrent neural network structure with temporal memory capabilities. Specifically, it can be implemented using a multi-layer network architecture including forget gates, input gates, and output gates. This model predicts future load status by capturing long-term dependencies in historical load characteristic data.

[0036] In this embodiment, standardized sensor data undergoes feature extraction to form a historical load feature dataset containing multidimensional time series data. This dataset is divided into a training set and a validation set for offline training of the long short-term memory prediction model. During deployment, standardized sensor data collected in real time undergoes the same feature extraction process and is input into the trained model for forward computation, outputting predicted load data for future time periods. The model adjusts the weights of hidden layer neurons to bring the mean squared error between the predicted results and the actual load data to a convergence threshold.

[0037] In this embodiment, the proposed solution models multidimensional time-series data using a long short-term memory (LSTM) prediction model. This model captures the periodicity, suddenness, and gradualness of load changes, significantly improving prediction accuracy. For example, in scenarios involving sudden server computing tasks, the model can predict load increases in advance based on historical vibration amplitude abrupt change patterns, achieving high-precision load prediction based on time-series characteristics. This allows fan speed control to respond to load changes in advance. The prediction results serve as input parameters for multi-objective optimization, effectively avoiding the problems of insufficient heat dissipation or excessive energy consumption caused by response lag in traditional methods, while also reducing the system stability risks caused by sudden load changes.

[0038] In one feasible implementation, refer to Figure 3 Step S400 includes steps S410 to S440, wherein: Step S410: Calculate the current load rate data based on the predicted load data; Step S420: Input the current load rate data, the preset regional temperature difference data, and the historical power consumption data into the multi-agent collaborative optimization model; Step S430: Optimization calculations are performed using the multi-agent collaborative optimization model, and the objective function value data is output. Step S440: Generate fan control command data based on the target function value data.

[0039] In this embodiment, the current load rate data refers to a real-time load status indicator determined by the ratio of predicted load data to the system's maximum carrying capacity. Specifically, it can be achieved by using a sliding window statistical method to calculate the ratio of the average load to the peak load within a period, reflecting the intensity of the server's current processing tasks. Regional temperature difference data refers to a quantitative indicator of temperature differences between different heat dissipation areas. Specifically, it can be achieved by collecting data from temperature sensors deployed at different locations on the server rack and calculating the absolute value of the temperature difference between adjacent areas, used to identify local areas with high heat dissipation demands. Historical power consumption data refers to fan energy consumption information recorded by the system in past operating cycles. Specifically, it can be achieved by storing data in a time-series database and querying by timestamp index, used to establish a correlation model between power consumption, load, and temperature. The multi-agent collaborative optimization model refers to a distributed optimization framework composed of multiple decision-making units. Specifically, it can be achieved by using reinforcement learning algorithms to train independent agents to control parameters such as fan speed and voltage frequency, used to coordinate conflicts between different control objectives. Objective function value data refers to the optimization result score calculated by comprehensively considering multiple dimensions such as energy consumption, noise, and heat dissipation efficiency. Specifically, it can be generated using weighted summation or Pareto front analysis methods, used to determine the optimal control strategy.

[0040] In this embodiment, the predicted load data is normalized and compared with the system's maximum performance threshold to generate the current load rate data. This data, along with real-time collected regional temperature difference data and power consumption data from historical operating cycles, constitutes the input parameters of the multi-agent collaborative optimization model. The independent decision-making units in the multi-agent collaborative optimization model perform local optimizations on heat dissipation efficiency, energy consumption control, and mechanical stability, generating a globally optimal solution through an interactive game mechanism. The objective function value data balances the priorities among different optimization objectives through a dynamic weight adjustment mechanism, and the final generated fan control command data includes the speed adjustment range, start / stop time interval, and collaborative working mode parameters.

[0041] Compared to existing technologies, traditional methods typically rely on a single temperature threshold to trigger a fixed speed control strategy, failing to dynamically coordinate the complex relationships between load changes, regional temperature differences, and historical energy consumption. This solution, however, utilizes a multi-agent model to achieve collaborative processing of multi-dimensional data, reducing redundant power consumption while ensuring heat dissipation efficiency and preventing equipment damage caused by localized overheating or excessive vibration. Through this technical solution, this application effectively addresses the response lag and target conflict issues present in traditional fan control strategies, achieving dynamic matching between load fluctuations and heat dissipation demands. By introducing a multi-dimensional data collaborative optimization mechanism, overall energy consumption is reduced while heat dissipation efficiency is improved, and mechanical vibration and noise caused by sudden speed changes are minimized.

[0042] In one feasible implementation, the step of performing optimization calculations and outputting objective function value data through the multi-agent collaborative optimization model includes: the fan control agent generating fan action data based on temperature and vibration data from the standardized sensor data; the DVFS control agent generating voltage-frequency combined action data based on load data and core temperature data; and jointly executing the fan action data and the voltage-frequency combined action data.

[0043] In this embodiment, the fan control agent refers to an independent decision-making module that dynamically adjusts the fan speed based on temperature and vibration data. Specifically, it can be implemented using a reinforcement learning-based policy network, generating corresponding speed adjustment commands by real-time sensing of heat dissipation needs and equipment vibration status. The DVFS control agent refers to an optimization module that dynamically adjusts the processor voltage and frequency according to changes in server load. Specifically, it can be implemented using a dynamic voltage and frequency adjustment algorithm, generating optimal power consumption configuration parameters by analyzing core temperature and computational workload. Fan action data includes fan speed adjustment commands, which can be implemented by driving the motor controller with pulse width modulation signals. Voltage-frequency combined action data includes the combined voltage and frequency parameters of the processor core, which can be dynamically configured by the motherboard power management unit. Joint execution refers to synchronously applying both types of action data to the server system, which can be achieved using a parallel control signal distribution mechanism to ensure the synergy between heat dissipation and power consumption adjustment.

[0044] In this embodiment, temperature and vibration data are input to the fan control agent, which calculates and generates fan speed adjustment commands via a policy network. For example, it increases the fan speed by a specific percentage when the temperature exceeds a set threshold. Load and core temperature data are input to the DVFS control agent, which generates voltage and frequency combination parameters using a dynamic voltage-frequency adjustment algorithm. For example, it reduces the core voltage to reduce power consumption under low load conditions. Subsequently, the fan action data and voltage-frequency combination action data are synchronously sent to the corresponding execution units. The fan module adjusts the airflow intensity according to the speed command, and the processor adjusts its operating state according to the voltage-frequency parameters. The two work together to achieve a balance between heat dissipation efficiency and energy consumption.

[0045] In this embodiment, the solution uses independent intelligent agents to handle fan control and processor frequency adjustment separately, and introduces a joint execution mechanism to achieve the goals of improving heat dissipation capacity and reducing power consumption simultaneously. This avoids the decrease in system stability caused by unilateral optimization and solves the energy efficiency imbalance problem caused by the separation of fan speed and processor frequency adjustment strategies in traditional methods. In high-load server scenarios, collaborative control avoids excessive core temperature, while reducing overall power consumption in low-load scenarios, thereby improving the dynamic response capability and energy utilization efficiency of the heat dissipation system.

[0046] In one feasible implementation, the step of jointly executing the fan action data and the voltage-frequency combined action data includes: after executing the fan action data and the voltage-frequency combined action data, collecting real-time power consumption data; collecting noise sensor data and converting it into linearized noise index data; calculating the standard deviation of vibration data to obtain stability index data; and combining the real-time power consumption data, linearized noise index data, and stability index data according to weights to calculate the objective function value data.

[0047] In this embodiment, real-time power consumption data refers to the actual electrical energy consumed after executing control actions. Specifically, it can be obtained by real-time acquisition of the voltage and current product using a power meter or current sensor, used to quantify energy consumption levels. Linearized noise index data refers to data that converts the nonlinear acoustic signal acquired by the sound pressure level sensor into a linear decibel value. This can be achieved through logarithmic transformation or A-weighted network processing, used for standardized noise assessment. Stability index data refers to statistical quantities reflecting the intensity of mechanical vibration. Specifically, it can be obtained by calculating the standard deviation or root mean square value of the vibration signal, used to characterize the smoothness of fan operation. Weighted combination refers to assigning proportional coefficients to different performance indicators and performing a weighted summation. Specifically, the weight coefficients can be determined using the analytic hierarchy process or entropy weighting method, used to balance conflicting relationships in multi-objective optimization.

[0048] In this embodiment, after generating the fan control command, the system simultaneously executes fan speed adjustment and processor voltage / frequency adjustment actions. Subsequently, it collects actual power consumption, noise, and vibration data through a sensor network. The noise signal undergoes linearization processing to eliminate frequency response deviation, and the vibration data is quantified using standard deviation calculation to determine mechanical stability. The real-time power consumption, linearized noise index, and stability index are weighted according to preset weights to obtain a comprehensive objective function value. This value is used to evaluate the merits of the current control strategy and guide subsequent optimization directions.

[0049] In this embodiment, by quantifying multi-dimensional operating parameters and using scientific methods to determine weights, a multi-objective dynamic balance is achieved, effectively resolving the contradiction between heat dissipation requirements under high load scenarios and energy saving and noise reduction under low load scenarios. Through the above technical solution, this application can monitor the comprehensive impact of fan control strategies on energy consumption, noise, and mechanical stability in real time, and dynamically evaluate the control effect through an objective function, ensuring that overall operating noise is reduced and equipment lifespan is extended while meeting heat dissipation requirements. This solution solves the problem of energy waste or equipment damage caused by neglecting the coupling relationship of multiple parameters in traditional methods, improving the reliability and environmental adaptability of server clusters.

[0050] In one feasible implementation, after the step of acquiring raw sensor data through the sensor group in the modular energy-saving array, the method further includes: performing spectral analysis processing on the vibration data in the raw sensor data to obtain vibration spectral characteristic data; generating a fault fan unit identification signal when the vibration spectral characteristic data exceeds a preset fault threshold; and triggering a hot-swap redundancy control mechanism based on the fault fan unit identification signal to activate the backup fan unit.

[0051] In this embodiment, spectrum analysis processing refers to the analysis method of converting vibration signals from the time domain to the frequency domain, specifically implemented using the Fast Fourier Transform algorithm, to identify abnormal frequency components in the vibration signal. Vibration spectrum feature data refers to the energy distribution data of each frequency band of the vibration signal extracted through spectrum analysis, specifically implemented using the frequency band energy ratio calculation method, to quantify the vibration intensity in different frequency ranges. The preset fault threshold refers to a pre-set standard for judging abnormal vibration spectrum, specifically derived from historical fault data statistics, to distinguish between normal vibration and abnormal vibration caused by mechanical faults. The faulty fan unit identification signal refers to a digital signal used to locate the faulty fan, specifically implemented using binary encoding, to accurately indicate the fan module that needs to be replaced. The hot-swappable redundancy control mechanism refers to the control logic for replacing hardware modules without power interruption, specifically implemented using dual power supply and bus isolation technology, to ensure continuous system operation during fault replacement. The backup fan unit refers to pre-deployed redundant heat dissipation equipment, specifically implemented using the same model of fan module, to quickly take over the heat dissipation task when a fault occurs.

[0052] In this embodiment, after the sensor group collects vibration data, the vibration signal is decomposed into energy distribution data of multiple frequency bands using a Fast Fourier Transform. For example, 0-100Hz is divided into the low-frequency band, 100-500Hz into the mid-frequency band, and above 500Hz into the high-frequency band, and the energy proportion of each frequency band is calculated. When the energy proportion of a certain frequency band exceeds a preset threshold, for example, the energy proportion of the mid-frequency band exceeds 60%, it is determined that the corresponding fan unit has a bearing wear or blade imbalance fault. At this time, an identification signal containing fault location information is generated, triggering the control circuit to cut off the power supply to the faulty fan, while activating the power supply line and communication interface of the backup fan. During this process, bus isolation technology ensures that the removal of the faulty module does not affect the data transmission of other fans, and the dual power supply design maintains the continuous power supply to the heat dissipation array.

[0053] Compared to existing technologies, traditional methods typically rely on manual inspections or single-threshold alarms, failing to accurately identify vibration anomalies and requiring downtime for maintenance during fault switching. This solution, however, utilizes spectral characteristic analysis to detect early mechanical faults, and its hot-swap mechanism reduces fault replacement time to milliseconds, avoiding the risk of server overheating due to maintenance downtime. Thus, this application can quickly locate the fault source and activate redundant cooling units when a fan experiences mechanical failure, effectively preventing localized overheating caused by a single fan failure, while reducing the frequency of manual intervention and improving the continuous operation and reliability of the server's cooling system.

[0054] In one feasible implementation, refer to Figure 4 Step S500 includes steps S510 to S540, wherein: Step S510: Train a local policy model at each edge node based on standardized sensor data to generate local model parameter data; Step S520: Add differential privacy noise to the local model parameter data to generate noise-added parameter data; Step S530: Homomorphic encryption is performed on the noise-added parameter data to generate encrypted model parameter data; Step S540: Aggregate multiple encryption model parameter data and perform policy update processing to generate global optimization policy data.

[0055] In this embodiment, the federated learning collaboration mechanism refers to a distributed machine learning framework. Specifically, it employs a method where edge nodes train models locally and then upload the parameters to the cloud for aggregation. Its purpose is to achieve multi-node collaborative training while protecting data privacy. Differential privacy noise involves adding random noise to the original data, specifically using Laplacian noise or Gaussian noise algorithms. Its function is to prevent sensitive information from being derived from local model parameters. Homomorphic encryption allows computation on the data in an encrypted state, specifically using Paillier encryption or fully homomorphic encryption schemes. Its function is to ensure that model parameters are not stolen or tampered with during transmission and aggregation.

[0056] In this embodiment, after standardized sensor data is distributed to each edge node, each node independently trains a local policy model based on its local data, generating corresponding local model parameters. Before uploading the parameters, differential privacy noise is injected into the model parameters to prevent them from being reverse-correlated to a specific node. Subsequently, the noise-added parameters are converted into ciphertext using a homomorphic encryption algorithm, ensuring that they cannot be decrypted even if intercepted during transmission. After receiving multiple encrypted model parameters, the cloud server performs aggregation operations, such as weighted averaging or gradient stacking, in the ciphertext state to update the global model parameters. The updated global optimization policy data is then distributed to each edge node to guide the next round of local model training iterations.

[0057] Compared to existing technologies, traditional multi-node collaborative control methods typically share raw data or plaintext parameters directly, posing a risk of data leakage and hindering cross-node collaborative optimization. Centralized training models in existing technologies are prone to creating data silos, and plaintext parameter transmission can be intercepted by malicious nodes. This solution utilizes a federated learning framework to ensure data remains local, combining differential privacy and homomorphic encryption techniques to effectively prevent sensitive information leakage and external attacks while maintaining the effectiveness of multi-node collaborative training.

[0058] Through the above technical solution, this application resolves the contradiction between data privacy and collaborative optimization in a multi-server cluster environment. While ensuring the security of sensor data at each node, it improves the dynamic control accuracy of the fan array by generating a global optimization strategy through encrypted aggregation. Simultaneously, the introduction of differential privacy noise prevents the reverse analysis of local model parameters, and homomorphic encryption ensures the security of data transmission, enabling the reliable implementation of the multi-node collaborative control mechanism in complex network environments.

[0059] In one feasible implementation, the method further includes: collecting external ambient temperature data; when the external ambient temperature data is lower than a low temperature threshold, controlling the fan to stop and activating the phase change material for heat dissipation; when the external ambient temperature data is higher than a high temperature threshold, activating redundant fan unit data and starting the liquid cooling auxiliary system.

[0060] In this embodiment, the external ambient temperature data refers to the real-time temperature information of the server room, which can be collected by temperature sensors deployed in the server room to reflect the temperature conditions of the server's operating environment. The low-temperature threshold is a preset critical temperature value that triggers the fan to stop, for example, 15°C. The specific value can be adjusted according to the server room design standards and is used to determine whether to activate passive cooling mode. The high-temperature threshold is a critical temperature value that triggers active cooling enhancement measures, for example, 35°C, used to identify the cooling needs in extreme high-temperature environments. Phase change material cooling refers to passive cooling by utilizing the heat absorption characteristics of phase change materials during phase change. Specifically, paraffin-based or metal-based phase change materials can be encapsulated in the server's heat dissipation module to replace fans for heat storage in low-temperature environments. Redundant fan unit data refers to pre-configured backup fan equipment information, which can be stored through the device management module to quickly activate additional cooling capacity at high temperatures. The liquid cooling auxiliary system is a liquid-circulation-based enhanced cooling device, specifically including a pump, coolant piping, and heat sink, used to improve cooling efficiency when the high-temperature threshold is triggered.

[0061] In this embodiment, external ambient temperature data is periodically collected by temperature sensors deployed in different areas of the server room and transmitted to the control module for threshold comparison. When the detected temperature is below the low-temperature threshold, the control module sends a stop command to the fan array and simultaneously sends an activation signal to the phase change material heat dissipation unit, causing the phase change material encapsulated in the heat sink to begin absorbing the heat generated by the server. When the detected temperature is above the high-temperature threshold, the control module first calls the redundant fan unit database, selects a suitable backup fan to start based on the current load, and simultaneously sends a start command to the liquid cooling system controller, causing the coolant to begin circulating and carrying away the heat accumulated inside the server.

[0062] In some specific implementations, the phase change material heat dissipation module can be designed with a replaceable slot structure, facilitating the replacement of materials with different phase change points according to the ambient temperature range of the computer room. The startup process of the liquid cooling auxiliary system may include a coolant pre-circulation stage, reducing system latency by establishing coolant flow paths in advance. The activation of redundant fan units can be combined with load forecast data to prioritize the activation of standby fans corresponding to high-load areas.

[0063] Compared to existing technologies, traditional solutions rely solely on a single fan speed control mechanism to address temperature changes, failing to effectively utilize passive cooling resources in low-temperature environments and lacking multi-level cooling enhancement measures in extreme high-temperature conditions. This solution introduces an ambient temperature threshold judgment mechanism to achieve intelligent switching between active and passive cooling modes. Simultaneously, it combines redundant fans with the coordinated control of the liquid cooling system to form a multi-level cooling guarantee system. Thus, this application solves the problems of significant energy waste in low-temperature environments and insufficient cooling capacity in extreme high-temperature environments associated with traditional fan control methods. Through an ambient temperature-triggered mode switching mechanism, energy consumption in low-load, low-temperature scenarios is significantly reduced while maintaining cooling performance. Simultaneously, in high-temperature scenarios, the coordinated control of the multi-level cooling system effectively prevents server overheating risks, improving the environmental adaptability and reliability of the cooling system.

[0064] In the embodiments of this application, the dynamic control method for modular energy-saving arrays of server fans achieves accurate prediction of load changes and optimization of heat dissipation resource allocation through multi-dimensional data acquisition and federated learning collaborative optimization, combined with dynamic speed regulation and redundancy control mechanisms. It has the advantages of reducing energy consumption, improving heat dissipation efficiency and equipment operation stability.

[0065] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the dynamic control method of the modular energy-saving array for server fans in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0066] This application also provides a dynamic control system for a modular energy-saving array of server fans. The dynamic control system for a modular energy-saving array of server fans includes: a memory 10, a processor 20, and a dynamic control program for a modular energy-saving array of server fans stored on the memory 10 and executable on the processor 20. The dynamic control program for a modular energy-saving array of server fans is configured to implement the steps of the dynamic control method for a modular energy-saving array of server fans.

[0067] The server fan modular energy-saving array dynamic control system provided in this application, employing the server fan modular energy-saving array dynamic control method in the above embodiments, can achieve precise speed adjustment to reduce energy consumption, improve heat dissipation efficiency, and enhance operational stability. Compared with the prior art, the beneficial effects of the server fan modular energy-saving array dynamic control system provided in this application are the same as those of the server fan modular energy-saving array dynamic control method provided in the above embodiments, and other technical features in the server fan modular energy-saving array dynamic control system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0068] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0069] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A dynamic control method for a modular energy-saving array of server fans, characterized in that, The method includes: Raw sensor data, including temperature data, vibration data, and airflow data, is collected through a sensor array in a modular energy-saving array. The raw sensor data is preprocessed to obtain standardized sensor data. The standardized sensor data is subjected to load prediction processing to obtain predicted load data; Based on the predicted load data, the preset regional temperature difference data, and the historical power consumption data, multi-objective dynamic speed regulation processing is performed to generate fan control command data. Through a federated learning collaboration mechanism, local model parameter data from multiple edge nodes are aggregated on a cloud server to generate global optimization strategy data. The final control signal is generated based on the global optimization strategy data and fan control command data to control the operating status of the fan modular array.

2. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, The steps for preprocessing the raw sensor data to obtain standardized sensor data include: The vibration data in the original sensor data is subjected to wavelet transform noise reduction processing to obtain noise-reduced vibration data. The temperature data and airflow data in the original sensor data are processed by Kalman filtering to obtain filtered temperature data and filtered airflow data. The noise-reduced vibration data, filtered temperature data, and filtered airflow data are normalized to obtain standardized sensor data.

3. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, The steps for performing load prediction processing on the standardized sensor data to obtain predicted load data include: Extract historical load characteristic data from the standardized sensor data; The historical load feature data is input into a pre-trained long short-term memory prediction model for time-series prediction processing, and the predicted load data is output.

4. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, Based on the predicted load data, preset regional temperature difference data, and historical power consumption data, the steps for performing multi-objective dynamic speed regulation processing and generating fan control command data include: Calculate the current load rate data based on the predicted load data; The current load rate data, the preset regional temperature difference data, and the historical power consumption data are input into the multi-agent collaborative optimization model. The optimization calculation is performed through the multi-agent collaborative optimization model, and the objective function value data is output. Fan control command data is generated based on the objective function value data.

5. The dynamic control method for modular energy-saving array of server fans as described in claim 4, characterized in that, The step of performing optimization calculations through the multi-agent collaborative optimization model and outputting objective function value data includes: The fan control agent generates fan action data based on the temperature and vibration data from the standardized sensor data. The DVFS control agent generates voltage-frequency combined action data based on load data and core temperature data; The fan action data and voltage-frequency combined action data are executed together.

6. The dynamic control method for modular energy-saving array of server fans as described in claim 5, characterized in that, The steps for jointly executing the fan action data and the voltage-frequency combined action data include: After executing the fan action data and voltage-frequency combined action data, real-time power consumption data is collected; Collect noise sensor data and convert it into linearized noise index data; The standard deviation of the vibration data is calculated to obtain the stability index data. The objective function value is calculated by combining the real-time power consumption data, linearized noise index data, and stability index data according to weights.

7. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, Following the step of acquiring raw sensor data via a sensor array in a modular energy-efficient array, the method further includes: The vibration data in the original sensor data is subjected to spectral analysis to obtain vibration spectral characteristic data; When the vibration spectrum characteristic data exceeds the preset fault threshold, a faulty fan unit identification signal is generated; The hot-swap redundancy control mechanism is triggered based on the faulty fan unit identification signal to activate the backup fan unit.

8. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, The steps for generating global optimization strategy data by aggregating local model parameter data from multiple edge nodes on a cloud server through a federated learning collaborative mechanism include: Local policy models are trained at each edge node based on standardized sensor data to generate local model parameter data. Differential privacy noise is added to the local model parameter data to generate noise-added parameter data; The noise-added parameter data is homomorphically encrypted to generate encrypted model parameter data. The system aggregates parameter data from multiple encryption models and performs policy update processing to generate global optimization policy data.

9. The dynamic control method for modular energy-saving array of server fans as described in claim 1, characterized in that, The method further includes: Collect external ambient temperature data; When the ambient temperature is below the low temperature threshold, the fan is stopped and the phase change material is activated to dissipate heat. When the external ambient temperature exceeds the high temperature threshold, the redundant fan unit data is activated and the liquid cooling auxiliary system is started.

10. A dynamic control system for a modular energy-saving array of server fans, characterized in that, The server fan modular energy-saving array dynamic control system includes: a memory, a processor, and a server fan modular energy-saving array dynamic control program stored in the memory and executable on the processor. The server fan modular energy-saving array dynamic control program is configured to implement the steps of the server fan modular energy-saving array dynamic control method as described in any one of claims 1 to 9.

Images