Cooling equipment control method and device, equipment and storage medium

Abstract

The invention discloses a heat dissipation equipment control method and device, equipment and a storage medium, and relates to the technical field of heat dissipation, and the method comprises the steps: collecting the actual operation feature data of heat dissipation equipment at a preset time interval, carrying out the feature processing of the actual operation feature data, obtaining the time sequence feature data of the heat dissipation equipment, and obtaining the time sequence feature data of the heat dissipation equipment; when a preset number of time sequence feature data is continuously collected, predicted operation feature data is obtained through a time sequence prediction model according to the time sequence feature data of the heat dissipation equipment, the predicted operation feature data and the time sequence feature data serve as input of a reinforcement learning model, and a control strategy of the heat dissipation equipment is generated; and the heat dissipation equipment is controlled according to the control strategy, so that the technical problem that the heat dissipation effect of the equipment load is reduced can be solved, and the heat dissipation effect is improved.

Description

Technical Field

[0001] This application relates to the technical field of heat dissipation, and particularly to a method, device, equipment and storage medium for controlling a heat dissipation device. Background Art

[0002] The heat dissipation system is a key component to ensure the stable operation of the device. The effectiveness of the control strategy of the heat dissipation system directly affects the hardware life, energy consumption and reliability of the device.

[0003] In the current related technologies, the heat dissipation system adopts a fixed threshold control strategy, that is, a preset air outlet temperature threshold is set. When the real-time temperature exceeds the threshold, the fan automatically runs at full speed. However, in the related technologies, the temperature of the device load is dynamically changing. For the changing load temperature with a fixed temperature threshold, problems of excessive heat dissipation performance and insufficient heat dissipation may occur, resulting in a reduction in the heat dissipation effect on the device load. Summary of the Invention

[0004] This application provides a method, device, equipment and storage medium for controlling a heat dissipation device, so as to at least solve the problem of reduced heat dissipation effect on the device load in the related technologies.

[0005] This application provides a method for controlling a heat dissipation device, including:

[0006] Collecting the actual operation characteristic data of the heat dissipation device at every preset time interval;

[0007] Performing feature processing on the actual operation characteristic data to obtain the time series characteristic data corresponding to the heat dissipation device;

[0008] When continuously obtaining a preset number of time series characteristic data, inputting the time series characteristic data into a trained time series prediction model, and outputting the predicted operation characteristic data of the heat dissipation device within a preset future time period;

[0009] Inputting the predicted operation characteristic data and the time series characteristic data into a trained reinforcement learning model, and outputting the control strategy of the heat dissipation device;

[0010] Controlling the heat dissipation device according to the control strategy.

[0011] This application also provides a device for controlling a heat dissipation device, including:

[0012] A collection module, configured to collect the actual operation characteristic data of the heat dissipation device at every preset time interval;

[0013] A feature processing module, configured to perform feature processing on the actual operation characteristic data to obtain the time series characteristic data corresponding to the heat dissipation device;

[0014] A first output module, configured to, when continuously obtaining a preset number of time series feature data, input the time series feature data into a trained time series prediction model, and output predicted operation feature data of the heat dissipation device within a preset future time period;

[0015] A second output module, configured to input the predicted operation feature data and the time series feature data into a trained reinforcement learning model, and output a control strategy for the heat dissipation device;

[0016] A control module, configured to control the heat dissipation device according to the control strategy.

[0017] The present application further provides an electronic device, including: a memory, configured to store a computer program; a processor, configured to implement the steps of any one of the above heat dissipation device control methods when executing the computer program.

[0018] The present application further provides a computer-readable storage medium, in which a computer program is stored, and wherein the computer program implements the steps of any one of the above heat dissipation device control methods when executed by a processor.

[0019] The present application further provides a computer program product, including a computer program, and the computer program implements the steps of any one of the above heat dissipation device control methods when executed by a processor.

[0020] Through the present application, since the actual operation feature data of the heat dissipation device is collected according to a preset time interval, the actual operation feature data is subjected to feature processing to obtain the time series feature data of the heat dissipation device. When continuously collecting a preset number of time series feature data, through the time series prediction model, the predicted operation feature data is obtained according to the time series feature data of the heat dissipation device, and the predicted operation feature data and the time series feature data are used as the input of the reinforcement learning model to generate a control strategy for the heat dissipation device, and the heat dissipation device is controlled according to the control strategy. Therefore, the technical problem of reducing the heat dissipation effect of the device load can be solved, and the heat dissipation effect is improved. Description of the Drawings

[0021] To more clearly illustrate the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a schematic system structure diagram of a computer device provided by an embodiment of the present application;

[0023] Figure 2 It is a schematic flowchart of a heat dissipation device control method provided by an embodiment of the present application;

[0024] Figure 3 This is a schematic structural diagram of the heat dissipation device control device provided by the embodiment of the present application;

[0025] Figure 4 This is a schematic structural diagram of the electronic device provided by the present application. Detailed implementation manners

[0026] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0027] It should be noted that in the description of the present application, the terms "including", "comprising" or any other variation thereof are intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. The terms "first", "second", etc. in the present application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0028] In order to solve the problem of reduced heat dissipation effect on the device load in the related art, the embodiments of the present application propose the following technical concept: The inventor considers collecting the actual operation characteristic data of the heat dissipation device at a preset time interval, considering performing feature processing on the actual operation characteristic data to obtain the time series characteristic data of the heat dissipation device. When a preset number of time series characteristic data are continuously collected, considering using a time series prediction model to predict and output the predicted operation characteristic data of the heat dissipation device, taking the predicted operation characteristic data and the time series characteristic data as the input of the reinforcement learning model, and generating a control strategy for the heat dissipation device through the reinforcement learning model to control the heat dissipation device, thereby improving the heat dissipation effect.

[0029] In order to enable those skilled in the art of the present technology to better understand the solution of the present application, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0030] Combined with the specific application environment architecture or specific hardware architecture on which the execution of the heat dissipation device control method depends, the specific application environment architecture or specific hardware architecture is described herein.

[0031] Refer to Figure 1 , Figure 1 This is a schematic system structure diagram of the computer device provided by the embodiment of the present application. As Figure 1As shown in the figure, the computer device includes: a receiving device 101, a processing device 102, and a display device 103.

[0032] It can be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the heat dissipation device control method. In some other feasible embodiments of the present application, the above architecture may include more or fewer components than those shown in the figure, or combine certain components, or split certain components, or have different component arrangements, which can be specifically determined according to the actual application scenario and will not be limited here. Figure 1 The components shown can be implemented in hardware, software, or a combination of software and hardware.

[0033] In the specific implementation process, the receiving device 101 can be an input / output interface or a communication interface, and can collect the actual operation characteristic data of the heat dissipation device.

[0034] The processing device 102 can generate a control strategy for the heat dissipation device.

[0035] The display device 103 can be used to display the control strategy of the above heat dissipation device, etc.

[0036] The display device can also be a touch display screen, which is used to receive user instructions while displaying the above content to realize the operation interaction with the user.

[0037] It should be understood that the above processing device can be implemented by a processor reading and executing instructions in a memory, or can be implemented by a chip circuit.

[0038] In addition, the network architecture and service scenarios described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those of ordinary skill in the art know that with the evolution of the network architecture and the emergence of new service scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0039] Figure 2 It is a schematic flowchart of the heat dissipation device control method provided by the embodiments of the present application. As Figure 2 shown, the embodiments of the present application provide a heat dissipation device control method, and the method will be described in detail as follows:

[0040] In this embodiment, the devices applying the heat dissipation device control method include but are not limited to: computers, servers, switches, and storage devices.

[0041] S201: Collect the actual operation characteristic data of the heat dissipation device at preset time intervals.

[0042] In this embodiment, the actual operating characteristic data of the heat dissipation device is collected by a data collection process.

[0043] In this embodiment, the preset time intervals of different heat dissipation devices are different.

[0044] In this embodiment, the actual operating characteristic data includes but is not limited to the inlet air temperature, the outlet air temperature, the overall power consumption, and the fan speed.

[0045] S202: Perform feature processing on the actual operating characteristic data to obtain the time series feature data corresponding to the heat dissipation device.

[0046] Specifically, calculate the temperature difference according to the outlet air temperature and the inlet air temperature, calculate the speed change rate according to the fan speed, calculate the power consumption change rate according to the overall power consumption, and perform normalization processing to obtain the time series feature data.

[0047] In this embodiment, the time series feature data includes but is not limited to the inlet air temperature, the outlet air temperature, the overall power consumption, the fan speed, the temperature difference information, the speed change rate, and the power consumption change rate.

[0048] S203: When a preset number of time series feature data is continuously obtained, input the time series feature data into the trained time series prediction model, and output the predicted operating characteristic data of the heat dissipation device within a preset future time period.

[0049] Specifically, select a sliding time window, obtain the continuous preset number of time series feature data within the sliding time window, input the time series feature data into the time series prediction model, and output the predicted operating characteristic data of the heat dissipation device within a preset future time period.

[0050] In this embodiment, the time series feature data of the previous 30 seconds before the current moment is collected.

[0051] In this embodiment, the predicted operating characteristic data is the predicted outlet air temperature at the current moment.

[0052] S204: Input the predicted operating characteristic data and the time series feature data into the trained reinforcement learning model, and output the control strategy of the heat dissipation device.

[0053] Specifically, input the predicted outlet air temperature and the time series feature data into the trained reinforcement learning model, and output the control strategy of the heat dissipation device.

[0054] S205: Control the heat dissipation device according to the control strategy.

[0055] In this embodiment, the control strategy includes but is not limited to the speed adjustment strategy of the heat dissipation device, the heat dissipation duration strategy, and the number of heat dissipation devices enabled.

[0056] As can be seen from the above embodiments, the actual operation characteristic data of the heat dissipation device is collected at a preset time interval, the actual operation characteristic data is subjected to feature processing to obtain the time series characteristic data of the heat dissipation device. When a preset number of time series characteristic data are continuously collected, the predicted operation characteristic data is obtained according to the time series characteristic data of the heat dissipation device through a time series prediction model. The predicted operation characteristic data and the time series characteristic data are used as the input of the reinforcement learning model to generate a control strategy for the heat dissipation device, and the heat dissipation device is controlled according to the control strategy, thereby improving the heat dissipation effect.

[0057] In an embodiment of the present application, step S202 includes:

[0058] S2021: Obtain the outlet temperature and the inlet temperature in the actual operation characteristic data, and calculate the temperature difference information according to the outlet temperature and the inlet temperature.

[0059] In this embodiment, the outlet temperature is denoted as .

[0060] In this embodiment, the inlet temperature is denoted as .

[0061] Specifically, the calculation method of the temperature difference information is:

[0062]

[0063] In the formula, represents the temperature difference at the (t - 1)th moment; represents the outlet temperature at the (t - 1)th moment; represents the inlet temperature at the (t - 1)th moment.

[0064] S2022: Obtain the fan speed in the actual operation characteristic data, and calculate the speed change rate according to the fan speed.

[0065] In this embodiment, the speed change rate is an index of the fluctuation degree of the fan speed in the time series, and is the standard deviation of the fan speed within a period of time.

[0066] In this embodiment, the speed change rate is denoted as .

[0067] In this embodiment, if the speed change rate is large, it means that the fan speed is frequently adjusted; if the speed change rate is small, it means that the fan speed is stable.

[0068] In this embodiment, a window length of 5 time steps is selected, and the standard deviation of the fan speed within the window length is recorded.

[0069] In this embodiment, 5 time steps cover the feedback period after the fan speed is adjusted.

[0070] In this embodiment, if the fan speed fluctuation is greater than 200 revolutions per minute and the adjustment frequency is greater than 3 times per minute, it is determined as high-frequency oscillation, triggering smooth control to limit the step size of a single adjustment.

[0071] Among them, the formula for the smooth control to adjust the speed is:

[0072]

[0073] In the formula, represents the adjusted speed; represents the speed before adjustment.

[0074] S2023: Obtain the overall power consumption in the actual operation characteristic data, and calculate the power consumption change rate based on the overall power consumption.

[0075] In this embodiment, the formula for calculating the power consumption change rate is:

[0076]

[0077] In the formula, represents the power consumption change rate; represents the overall power consumption at the current moment; represents the overall power consumption at an interval from the time T with a sampling interval; represents the sampling interval, with a duration of 1 second.

[0078] In this embodiment, when it is detected that the power consumption change rate is greater than 100 W / s, it is determined that the load has increased suddenly, triggering the time series prediction model to predict the temperature change in the next 30 seconds.

[0079] In this embodiment, when the standard deviation of the power consumption change rate, that is, the load volatility, is greater than 50 W / s, the time window is shortened, adjusted from 30 seconds to 20 seconds.

[0080] S2024: Normalize the temperature difference information, speed change rate, power consumption change rate, and actual operation characteristic data to obtain the time series characteristic data of the heat dissipation device.

[0081] In this embodiment, the minimum-maximum normalization method is used to normalize the data.

[0082] Among them, the formula for the normalization process is:

[0083]

[0084] In the formula, represents the unnormalized data; represents the minimum value of the unnormalized data; represents the maximum value of the unnormalized data; represents the normalized data.

[0085] As can be seen from the above embodiments, by obtaining the outlet temperature and the inlet temperature, calculating the temperature difference information, by obtaining the fan speed, calculating the speed change rate, by obtaining the overall power consumption, calculating the power consumption change rate, normalizing the temperature difference information, the speed change rate, the power consumption change rate and the actual operation characteristic data, constructing the time series characteristic data, and using it as the input of the time series prediction model, the singularity of the source of the operation characteristic data of the heat dissipation device is avoided.

[0086] In an embodiment of the present application, before step S204, it further includes:

[0087] S301: Obtain the predicted operation characteristic data of the heat dissipation device and the training set state vector of the time series characteristic data.

[0088] In this embodiment, the deep Q-network algorithm is used as the reinforcement learning algorithm.

[0089] Among them, the deep Q-network is a model-free reinforcement learning algorithm, which predicts the expected return of taking a certain action in a certain state by learning a state-action value function.

[0090] S302: Construct a heat dissipation action according to the predicted operation characteristic data of the heat dissipation device and the training set state vector of the time series characteristic data.

[0091] In this embodiment, the heat dissipation action is the adjustment step of the fan speed.

[0092] Among them, the adjustment range is ±5% to ±20%.

[0093] Specifically, the adjustment step of the fan speed is discretized.

[0094] Exemplarily, the adjustment steps are divided into -20%, -15%, -10%, -5%, 0, +5%, +10%, +15% and +20%.

[0095] S303: Execute the heat dissipation action and generate a reward value corresponding to the heat dissipation action according to the reward function.

[0096] In this embodiment, the reward function is the feedback provided by the actions taken by the agent in different states.

[0097] In this embodiment, the formula for generating the reward value corresponding to the heat dissipation action according to the reward function is:

[0098]

[0099] In the formula, represents the reward value; represents the predicted air outlet temperature; represents the highest temperature threshold for the safe operation of the server; represents the adjusted rotational speed; represents the rotational speed before adjustment; represents the adjusted power consumption; represents the power consumption before adjustment.

[0100] S304: Store the training set state vectors, cooling actions, and the reward values corresponding to the cooling actions of the predicted operating characteristic data and time series characteristic data of the cooling device in the experience pool of the reinforcement learning model.

[0101] Specifically, optimize the cooling action according to the magnitude of the reward value, and use it as the experience for the next round of reinforcement learning to optimize the cooling action in the next round.

[0102] S305: Update the state vectors, cooling actions, and reward values in the experience pool according to the preset number of iterations to generate updated reinforcement learning model parameters.

[0103] Specifically, perform iterative learning according to the preset number of iterations to obtain the cooling action with the maximum reward value, and obtain the updated reinforcement learning model parameters.

[0104] S306: Determine the reinforcement learning model according to the updated reinforcement learning model parameters.

[0105] Specifically, if the updated reinforcement learning model parameters meet the training stop policy of the reinforcement learning, determine the reinforcement learning model according to the reinforcement learning model parameters.

[0106] As can be seen from the above embodiments, by obtaining the training set state vectors, constructing the cooling actions, calculating the reward values for performing the cooling actions under the conditions of the training set state vectors, storing the state vectors, cooling actions, and reward values in the experience pool of the reinforcement learning model, performing iterative learning, and training to obtain the reinforcement learning model, and formulating the cooling strategy through the reinforcement learning model, the accuracy of cooling is improved.

[0107] In an embodiment of the present application, after step S306, it further includes:

[0108] S307: Obtain the device specification information of the cooling device.

[0109] In this embodiment, the device specification information includes but is not limited to the number of operating devices, device power consumption, and device type.

[0110] S308: Determine the cooling type according to the device specification information of the cooling device.

[0111] In this embodiment, the heat dissipation types include but are not limited to high heat dissipation demand type, energy-saving type, and low mechanical stress type.

[0112] S309: Adjust the temperature deviation penalty weight, rotation speed change penalty weight, and power consumption fluctuation penalty weight in the reward function according to the heat dissipation type to adapt to the heat dissipation type.

[0113] In this embodiment, the temperature deviation penalty measures the deviation degree between the predicted temperature and the set safety temperature.

[0114] In this embodiment, the rotation speed change penalty suppresses the frequent fluctuation of the fan rotation speed, reducing mechanical loss and noise.

[0115] In this embodiment, the power consumption fluctuation penalty is the additional power consumption caused by the rotation speed adjustment.

[0116] In this embodiment, the temperature deviation penalty is a high-priority target, the rotation speed change penalty is a medium-priority target, and the power consumption fluctuation penalty is a low-priority target.

[0117] Exemplarily, for devices of the low mechanical stress type, increase the rotation speed change penalty weight and reduce the fan rotation speed to adapt to the device.

[0118] As can be seen from the above embodiments, by obtaining the device specification information of the heat dissipation device, determining the heat dissipation type, and adjusting the temperature deviation penalty weight, rotation speed change penalty weight, and power consumption fluctuation penalty weight in the reward function according to different heat dissipation types to adapt to different types of heat dissipation devices, it is convenient for the reinforcement learning model to generate a more accurate heat dissipation strategy.

[0119] In an embodiment of the present application, before step S203, it further includes:

[0120] S401: Obtain the training data set of the time series feature data of the heat dissipation device.

[0121] In this embodiment, the training data set includes but is not limited to the outlet temperature at historical moments, the inlet temperature at historical moments, the fan rotation speed at historical moments, and the overall machine power consumption at historical moments.

[0122] S402: Create a time series prediction model of the heat dissipation device.

[0123] In this embodiment, the time series prediction model includes three layers of long short-term memory network layers and a fully connected layer.

[0124] S403: Input the training data set of the time series feature data of the heat dissipation device into the long short-term memory network layer of the time series prediction model, and output the hidden features.

[0125] In this embodiment, the first long short-term memory network layer includes 64 neurons, and the result of the returned parameter is true. The second long short-term memory network layer includes 32 neurons, and the result of the returned parameter is true. The third long short-term memory network layer includes 16 neurons, and the result of the returned parameter is false.

[0126] In this embodiment, the returned parameter is denoted as return_sequences.

[0127] Among them, if return_sequences = true, it means that the returned parameter returns the output of each time step; if return_sequences = false, it means that the returned parameter only returns the output of the last time step.

[0128] S404: Input the hidden features into the fully connected layer of the time series prediction model, and output the predicted operating feature data corresponding to the training data set.

[0129] In this embodiment, the fully connected layer includes 1 neuron.

[0130] Specifically, perform inverse normalization processing on the output of the fully connected layer to obtain the predicted temperature value of the air outlet.

[0131] S405: If the predicted operating feature data corresponding to the training data set meets the preset early stopping strategy, determine the time series prediction model according to the predicted operating feature data corresponding to the training data set.

[0132] Specifically, calculate the mean square error of the validation data set. If the mean square error of the validation data set does not decrease for 10 consecutive rounds, the model parameters determined by the training data set meet the early stopping strategy. Obtain the corresponding model parameters according to the predicted operating feature data output by the model, and determine the model parameters of the time series prediction model that has completed training.

[0133] In this embodiment, an edge-cloud hierarchical control architecture is adopted. The time series prediction model runs through the edge node, the data of the time series prediction model is aggregated through the cloud, and the update strategy is sent down through the cloud.

[0134] As can be seen from the above embodiments, by obtaining the training data set, creating a time series prediction model for the heat dissipation device, inputting the training data set into the long short-term memory network layer, outputting hidden features, and outputting the predicted operating feature data corresponding to the training data set through the fully connected layer according to the hidden features, it is determined whether the predicted operating feature data meets the early stopping strategy to determine the time series prediction model, and the operating feature data is predicted in advance through the time series prediction model, providing a data basis for formulating the heat dissipation strategy.

[0135] In an embodiment of the present application, after step S405, the following steps are further included:

[0136] S406: If the predicted operation characteristic data corresponding to the training data set does not meet the preset early stopping strategy, set the initial learning rate and the number of iterations of the optimizer.

[0137] In this embodiment, the optimizer is selected as the Adam optimizer, the initial learning rate is set to 0.001, and the number of iterations is set to 100.

[0138] S407: Adjust the model parameters of the time series prediction model according to the initial learning rate and the number of iterations to obtain at least one set of predicted model parameters.

[0139] In this embodiment, the batch size is set to 32.

[0140] In this embodiment, the batch size is the number of samples used to calculate the gradient and update the model parameters.

[0141] S408: Apply at least one set of predicted model parameters to the time series prediction model to obtain at least one set of predicted operation characteristic data.

[0142] Specifically, apply the model parameters to the time series prediction model, and predict the time series characteristic data through the time series prediction model to obtain the predicted operation characteristic data corresponding to the output of each model parameter.

[0143] S409: Calculate the mean square error between at least one set of predicted operation characteristic data and the validation data set of the time series characteristic data of the heat dissipation device.

[0144] In this embodiment, the formula for calculating the mean square error is:

[0145]

[0146] In the formula, represents the mean square error; represents the actual outlet temperature; represents the predicted outlet temperature; N represents the number of training samples.

[0147] S410: If the mean square error meets the preset early stopping strategy, determine the time series prediction model according to at least one set of predicted operation characteristic data.

[0148] In this embodiment, the preset early stopping strategy is the mean square error of the validation set. When the mean square error of the validation set does not decrease for 10 consecutive rounds, the training is automatically stopped.

[0149] Specifically, if the mean square error of the validation set does not decrease for 10 consecutive rounds, save the model parameters corresponding to the minimum value of the average error of the validation set and determine the time series prediction model.

[0150] As can be seen from the above embodiments, if the predicted operation characteristic data does not meet the early stopping strategy, the initial learning rate and the number of iterations are set, the model parameters are adjusted, and the mean square error with the validation data set is calculated based on the predicted operation characteristic data output by the obtained multiple model parameters. If the mean square error meets the early stopping strategy, it is determined as the time series prediction model, and by calculating the mean square error of the validation data set, the phenomenon of overfitting in the training model is prevented.

[0151] In an embodiment of the present application, before step S203, it further includes:

[0152] S501: Obtain the historical rotation speed adjustment data set and the historical temperature change delay data set of the heat dissipation device.

[0153] In this embodiment, the historical temperature change delay is the delay of the temperature change caused by the heat dissipation device adjusting its rotation speed according to the rotation speed adjustment strategy.

[0154] S502: Establish a mapping relationship between the rotation speed and the temperature change according to the historical rotation speed adjustment data set and the historical temperature change delay data set.

[0155] In this embodiment, feature extraction is performed on the historical rotation speed and temperature sequences through a convolutional neural network to identify the delay distribution corresponding to different rotation speed change amplitudes, and a mapping relationship is established.

[0156] S503: Generate compensation data for the time series prediction model according to the mapping relationship between the rotation speed and the temperature change.

[0157] In this embodiment, the compensation data is used to compensate the prediction of the outlet temperature by the time series prediction model.

[0158] Exemplarily, when the rotation speed changes by 20%, the delay of the temperature drop is 8 - 12 seconds. By introducing a delay compensation time point of 10 seconds, the prediction accuracy of the outlet temperature by the time series prediction model is improved.

[0159] As can be seen from the above embodiments, by obtaining the historical data sets of the rotation speed adjustment and the temperature change delay, establishing a mapping relationship between the rotation speed and the temperature change, and calculating the compensation of the time series prediction model according to the mapping relationship, the situation where there is a physical delay in the temperature response after the fan rotation speed is adjusted is avoided, and the accuracy of predicting the outlet temperature is improved.

[0160] In an embodiment of the present application, after step S205, it further includes:

[0161] S206: Collect the operation characteristic data of the heat dissipation device after implementing the control strategy.

[0162] Specifically, after the execution control strategy is executed, the length of the set time window is set, and the operation characteristic data of the heat dissipation device within the time window is collected.

[0163] S207: Generate a temperature curve of the heat dissipation device according to the operation characteristic data of the heat dissipation device after the execution control strategy.

[0164] In this embodiment, the abscissa of the temperature curve of the heat dissipation device is the sampling time within the time window, and the ordinate is the temperature value of the heat dissipation device.

[0165] S208: If the temperature curve of the heat dissipation device exceeds the temperature curve threshold of the heat dissipation device, generate an alarm message.

[0166] Specifically, compare the drawn temperature curve of the heat dissipation device with the temperature curve threshold of the heat dissipation device, obtain the sampling points that exceed the temperature threshold, obtain the temperature and status information of the corresponding sampling points, and record and generate an alarm message.

[0167] As can be seen from the above embodiments, by collecting the operation characteristic data of the heat dissipation device after the execution control strategy, drawing a temperature curve, comparing the drawn temperature curve with the temperature curve threshold of the heat dissipation device, generating an alarm message for the part that exceeds the threshold, and prompting the staff to manually adjust the heat dissipation strategy.

[0168] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method.

[0169] Figure 3 It is a schematic structural diagram of the heat dissipation device control device provided by the embodiment of the present application. As Figure 3 shown, the embodiment of the present application also provides a heat dissipation device control device 30, including: a collection module 301, a feature processing module 302, a first output module 303, a second output module 304, and a control module 305.

[0170] The collection module 301 is used to collect the actual operation characteristic data of the heat dissipation device at every preset time interval;

[0171] The feature processing module 302 is used to perform feature processing on the actual operation characteristic data to obtain the time series feature data corresponding to the heat dissipation device;

[0172] The first output module 303 is used to input the time series feature data into the trained time series prediction model when continuously obtaining a preset number of time series feature data, and output the predicted operation characteristic data of the heat dissipation device within a preset future time period;

[0173] A second output module 304, configured to input the predicted operation feature data and the time series feature data into a trained reinforcement learning model, and output a control strategy for the heat dissipation device;

[0174] A control module 305, configured to control the heat dissipation device according to the control strategy.

[0175] In an embodiment of the present application, the feature processing module 302 includes:

[0176] A first acquisition unit, configured to acquire the air outlet temperature and the air inlet temperature in the actual operation feature data, and calculate temperature difference information according to the air outlet temperature and the air inlet temperature.

[0177] A second acquisition unit, configured to acquire the fan speed in the actual operation feature data, and calculate a speed change rate according to the fan speed.

[0178] A third acquisition unit, configured to acquire the overall power consumption in the actual operation feature data, and calculate a power consumption change rate according to the overall power consumption.

[0179] A normalization processing unit, configured to perform normalization processing on the temperature difference information, the speed change rate, the power consumption change rate, and the actual operation feature data to obtain the time series feature data of the heat dissipation device.

[0180] In an embodiment of the present application, the heat dissipation device control device 30 further includes:

[0181] A first acquisition module, configured to acquire the training set state vector of the predicted operation feature data and the time series feature data of the heat dissipation device.

[0182] A construction module, configured to construct a heat dissipation action according to the training set state vector of the predicted operation feature data and the time series feature data of the heat dissipation device.

[0183] An execution module, configured to execute the heat dissipation action and generate a reward value corresponding to the heat dissipation action according to the reward function.

[0184] A storage module, configured to store the training set state vector of the predicted operation feature data and the time series feature data of the heat dissipation device, the heat dissipation action, and the reward value corresponding to the heat dissipation action into the experience pool of the reinforcement learning model.

[0185] A first generation module, configured to update the state vector, the heat dissipation action, and the reward value in the experience pool according to a preset number of iterations, and generate updated reinforcement learning model parameters.

[0186] A first determination module, configured to determine the reinforcement learning model according to the updated reinforcement learning model parameters.

[0187] In an embodiment of the present application, the heat dissipation device control device 30 further includes:

[0188] A second acquisition module, configured to acquire the device specification information of the heat dissipation device.

[0189] A second determination module, configured to determine the heat dissipation type according to the device specification information of the heat dissipation device.

[0190] A first adjustment module, configured to adjust the temperature deviation penalty weight, the rotation speed change penalty weight, and the power consumption fluctuation penalty weight in the reward function according to the heat dissipation type, so as to adapt to the heat dissipation type.

[0191] In an embodiment of the present application, the heat dissipation device control device 30 further includes:

[0192] A third acquisition module, configured to acquire a training data set of the time series feature data of the heat dissipation device.

[0193] A creation module, configured to create a time series prediction model of the heat dissipation device.

[0194] A third output module, configured to input the training data set of the time series feature data of the heat dissipation device into the long short-term memory network layer of the time series prediction model, and output hidden features.

[0195] A fourth output module, configured to input the hidden features into the fully connected layer of the time series prediction model, and output the predicted operation feature data corresponding to the training data set.

[0196] A third determination module, configured to determine the time series prediction model according to the predicted operation feature data corresponding to the training data set if the predicted operation feature data corresponding to the training data set meets the preset early stopping strategy.

[0197] In an embodiment of the present application, the heat dissipation device control device 30 further includes:

[0198] A setting module, configured to set the initial learning rate and the number of iterations of the optimizer if the predicted operation feature data corresponding to the training data set does not meet the preset early stopping strategy.

[0199] A second adjustment module, configured to adjust the model parameters of the time series prediction model according to the initial learning rate and the number of iterations, so as to obtain at least one set of predicted model parameters.

[0200] An application module, configured to apply at least one set of predicted model parameters to the time series prediction model, so as to obtain at least one set of predicted operation feature data.

[0201] A calculation module, configured to calculate the mean square error between at least one set of predicted operation feature data and the verification data set of the time series feature data of the heat dissipation device.

[0202] A fourth determination module, configured to determine a time series prediction model according to at least one set of predicted operation characteristic data if the mean square error meets a preset early stopping strategy.

[0203] In an embodiment of the present application, the heat dissipation device control device 30 further includes:

[0204] A fourth acquisition module, configured to acquire a historical rotation speed adjustment data set and a historical temperature change delay data set of the heat dissipation device.

[0205] A building module, configured to establish a mapping relationship between the rotation speed and the temperature change according to the historical rotation speed adjustment data set and the historical temperature change delay data set.

[0206] A second generation module, configured to generate compensation data for the time series prediction model according to the mapping relationship between the rotation speed and the temperature change.

[0207] For the description of the features in the corresponding embodiment of the heat dissipation device control device, reference may be made to the relevant description of the corresponding embodiment of the heat dissipation device control method, which will not be elaborated here one by one.

[0208] Figure 4 It is a schematic structural diagram of an electronic device provided by the present application. As Figure 4 shown, the electronic device 40 provided in this embodiment includes: at least one processor 401 and a memory 402. Optionally, the electronic device 40 further includes a communication component 403. Among them, the processor 401, the memory 402, and the communication component 403 are connected through a bus.

[0209] In a specific implementation process, at least one processor 401 executes computer execution instructions stored in the memory 402, so that at least one processor 401 executes the above-mentioned heat dissipation device control method embodiment.

[0210] For the specific implementation process of the processor 401, reference may be made to the above method embodiment, and its implementation principle and technical effects are similar, which will not be elaborated here in this embodiment.

[0211] In the above embodiment, it should be understood that the processor may be a central processing unit (Central Processing Unit, abbreviated as: CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, abbreviated as: DSP), application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the application can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0212] The memory may include a random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0213] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0214] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored, and the computer program is configured to execute the steps in any one of the above embodiments of the heat dissipation device control method when running.

[0215] In an exemplary embodiment, the above computer-readable storage medium may include, but is not limited to: various media such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disc that can store a computer program.

[0216] An embodiment of the present application also provides a computer program product, the above computer program product includes a computer program, and the steps in any one of the above embodiments of the heat dissipation device control method are implemented when the computer program is executed by a processor.

[0217] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium stores a computer program, and the steps in any one of the above embodiments of the heat dissipation device control method are implemented when the computer program is executed by a processor.

[0218] Those skilled in the art may further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0219] The above has introduced in detail a heat dissipation device control method, device, equipment, and storage medium provided by this application. Specific examples are used herein to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of this application, several improvements and modifications can be made to this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for controlling a heat dissipation device, characterized in that, Including: Collecting the actual operation characteristic data of the heat dissipation device once every preset time interval; Performing feature processing on the actual operation characteristic data to obtain the time series characteristic data corresponding to the heat dissipation device; When continuously obtaining a preset number of time series characteristic data, inputting the time series characteristic data into a trained time series prediction model, and outputting the predicted operation characteristic data of the heat dissipation device within a preset future time period; Inputting the predicted operation characteristic data and the time series characteristic data into a trained reinforcement learning model, and outputting the control strategy of the heat dissipation device; Controlling the heat dissipation device according to the control strategy.

2. The heat dissipation device control method according to claim 1, characterized in that The performing feature processing on the actual operation characteristic data to obtain the time series characteristic data corresponding to the heat dissipation device includes: Obtaining the outlet temperature and the inlet temperature in the actual operation characteristic data, and calculating the temperature difference information according to the outlet temperature and the inlet temperature; Obtaining the fan speed in the actual operation characteristic data, and calculating the speed change rate according to the fan speed; Obtaining the overall power consumption in the actual operation characteristic data, and calculating the power consumption change rate according to the overall power consumption; Normalizing the temperature difference information, the speed change rate, the power consumption change rate and the actual operation characteristic data to obtain the time series characteristic data of the heat dissipation device.

3. The heat dissipation device control method according to claim 1, wherein, Before inputting the predicted operation characteristic data and the time series characteristic data into a trained reinforcement learning model, further including: Obtaining the training set state vector of the predicted operation characteristic data and the time series characteristic data of the heat dissipation device; Constructing a heat dissipation action according to the training set state vector of the predicted operation characteristic data and the time series characteristic data of the heat dissipation device; Performing the heat dissipation action, and generating a reward value corresponding to the heat dissipation action according to the reward function; Storing the training set state vector of the predicted operation characteristic data and the time series characteristic data of the heat dissipation device, the heat dissipation action and the reward value corresponding to the heat dissipation action into the experience pool of the reinforcement learning model; Updating the state vector, the heat dissipation action and the reward value in the experience pool according to a preset number of iterations, and generating updated reinforcement learning model parameters; Determining the reinforcement learning model according to the updated reinforcement learning model parameters.

4. The heat dissipation device control method according to claim 3, characterized in that The formula for generating the reward value corresponding to the heat dissipation action according to the reward function is: In the formula, represents the reward value; represents the predicted air outlet temperature; represents the highest temperature threshold for the safe operation of the server; represents the adjusted rotational speed; represents the rotational speed before adjustment; represents the adjusted power consumption; represents the power consumption before adjustment.

5. The heat dissipation device control method according to claim 3, wherein After determining the reinforcement learning model according to the updated reinforcement learning model parameters, further including: Obtaining the device specification information of the heat dissipation device; Determining the heat dissipation type according to the device specification information of the heat dissipation device; Adjusting the temperature deviation penalty weight, the speed change penalty weight and the power consumption fluctuation penalty weight in the reward function according to the heat dissipation type to adapt to the heat dissipation type.

6. The heat dissipation device control method according to claim 1, wherein, Before inputting the time series characteristic data into a trained time series prediction model, further including: Obtaining the training data set of the time series characteristic data of the heat dissipation device; Creating a time series prediction model of the heat dissipation device; Input the training data set of the time series feature data of the heat dissipation device into the long short-term memory network layer of the time series prediction model to output hidden features; Input the hidden features into the fully connected layer of the time series prediction model to output the predicted operating feature data corresponding to the training data set; If the predicted operating feature data corresponding to the training data set meets the preset early stopping strategy, determine the time series prediction model according to the predicted operating feature data corresponding to the training data set.

7. The heat dissipation device control method according to claim 6, wherein After determining the time series prediction model according to the predicted operating feature data corresponding to the training data set, it further includes: If the predicted operating feature data corresponding to the training data set does not meet the preset early stopping strategy, set the initial learning rate and the number of iterations of the optimizer; Adjust the model parameters of the time series prediction model according to the initial learning rate and the number of iterations to obtain at least one set of predicted model parameters; Apply the at least one set of predicted model parameters to the time series prediction model to obtain at least one set of predicted operating feature data; Calculate the mean square error between the at least one set of predicted operating feature data and the validation data set of the time series feature data of the heat dissipation device; If the mean square error meets the preset early stopping strategy, determine the time series prediction model according to the at least one set of predicted operating feature data.

8. The heat dissipation device control method according to claim 1, wherein Before inputting the time series feature data into the time series prediction model that has been trained, it further includes: Obtain the historical rotational speed adjustment data set and the historical temperature change delay data set of the heat dissipation device; Establish a mapping relationship between the rotational speed and the temperature change according to the historical rotational speed adjustment data set and the historical temperature change delay data set; Generate compensation data for the time series prediction model according to the mapping relationship between the rotational speed and the temperature change.

9. A heat dissipation device control device, characterized in that, It includes: A collection module for collecting the actual operating feature data of the heat dissipation device at every preset time interval; A feature processing module for performing feature processing on the actual operating feature data to obtain the time series feature data corresponding to the heat dissipation device; A first output module for, when continuously obtaining a preset number of time series feature data, inputting the time series feature data into the time series prediction model that has been trained and outputting the predicted operating feature data of the heat dissipation device within a preset future time period; A second output module for inputting the predicted operating feature data and the time series feature data into the trained reinforcement learning model and outputting the control strategy of the heat dissipation device; A control module for controlling the heat dissipation device according to the control strategy.

10. An electronic device, characterized in that, It includes: A memory for storing computer programs; A processor for implementing the steps of the heat dissipation device control method according to any one of claims 1 to 8 when executing the computer program.

11. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, wherein the computer program implements the steps of the heat dissipation device control method according to any one of claims 1 to 8 when executed by a processor.

Images