Computer cooling system based on deep reinforcement learning

Abstract

The invention discloses a computer heat dissipation system based on deep reinforcement learning, and the system comprises an air cooling subsystem which comprises an air inlet fan, a plurality of air outlet fans, and a plurality of ventilation holes; the water cooling subsystem comprises a heat dissipation box and a refrigeration box which are communicated end to end; an air guide scoop; a water cooling pipe; at least one semiconductor chilling plate; a pump; a radiating fin group I; a radiating fin group II; a plurality of temperature sensors; a plurality of rotating speed sensors; a current sensor; the control subsystem comprises a deep reinforcement learning module; and a control module. According to the computer cooling system based on deep reinforcement learning, the air cooling subsystem, the water cooling subsystem and the semiconductor chilling plate are combined, the deep reinforcement learning module conducts deep analysis and learning on collected and normalized data and continuously interacts with the system environment, the cooling strategy is continuously optimized, and the cooling efficiency is improved. And the most adaptive heat dissipation control instruction is output to the control module, so that the computer heat dissipation system is always in an efficient and intelligent running state.

Description

Technical Field

[0001] The present invention relates to the technical field of computer heat dissipation, and particularly relates to a computer heat dissipation system based on deep reinforcement learning. Background Art

[0002] The CPU and various electronic components inside a computer host generate a lot of heat during use. To ensure the normal operation of the computer, it is necessary to promptly dissipate the heat generated inside it. Moreover, with the continuous improvement of computer performance, the computing power of core components such as CPUs and GPUs is getting stronger and stronger, and the heat they generate is also increasing day by day, posing higher requirements for the heat dissipation efficiency of the heat dissipation system. The existing computer heat dissipation methods mostly use water cooling or air cooling. As a low-cost heat dissipation method, air cooling is widely used, but the cooling efficiency of the fan is difficult to change with the change of the CPU temperature. The water cooling heat dissipation system relies on the high specific heat capacity of water to take away heat through circulating water, and the heat dissipation efficiency is relatively high. However, there is a hidden danger of water leakage in the structure of the water cooling system. In addition, most of the existing heat dissipation systems implement heat dissipation control using pre-set fixed strategies, and cannot dynamically adjust according to the real-time workload and temperature changes of the computer, resulting in a situation of waste of heat dissipation resources or poor heat dissipation effect. Summary of the Invention

[0003] An object of the present invention is to solve at least the above problems and provide at least the advantages described later.

[0004] Another object of the present invention is to provide a computer heat dissipation system based on deep reinforcement learning, which combines an air cooling subsystem, a water cooling subsystem and a thermoelectric cooler. The deep reinforcement learning module continuously interacts with the system environment by deeply analyzing and learning the collected and normalized data, and continuously optimizes the heat dissipation strategy, and outputs the most suitable heat dissipation control instruction for the control module, so as to ensure that the computer heat dissipation system is always in an efficient and intelligent operating state.

[0005] To achieve these and other advantages of the present invention, a computer heat dissipation system based on deep reinforcement learning is provided, including: An air cooling subsystem, which includes an intake fan disposed in the middle of the front cover of the computer chassis; a plurality of exhaust fans respectively disposed on the side cover and the rear cover of the chassis; and a plurality of ventilation holes uniformly distributed on the chassis and disposed close to the intake fan and the plurality of exhaust fans; A water cooling subsystem, which includes a heat dissipation box and a refrigeration box connected end to end, and both the heat dissipation box and the refrigeration box are attached and disposed ... ... On the inner side wall of the front cover of the chassis, the heat dissipation box and the refrigeration box are respectively located above and below the intake fan; a wind guiding hopper, which is a funnel-shaped cylinder with a double-layer structure, the wide mouth end of the wind guiding hopper is arranged facing the intake fan and the refrigeration box, and the narrow mouth end of the wind guiding hopper extends to near the heat generating components of the computer; a water cooling pipe, one end of which is connected to the liquid outlet end of the refrigeration box, the main part of the water cooling pipe is coiled inside the double-layer structure of the wind guiding hopper, and the other end of the water cooling pipe is connected to the water inlet end of the heat dissipation box; a pump, which is arranged on the water cooling pipe; at least one thermoelectric cooler, the refrigerating surface of which is attached to the side wall of the refrigeration box, and the heat generating surface of the thermoelectric cooler is arranged facing the outside of the chassis through the air vent on the front cover; a heat sink group Ⅰ, which is arranged on the front cover of the chassis, and the heat sink group Ⅰ abuts against the outside of the heat dissipation box; a heat sink group Ⅱ, which is arranged on both sides of the chassis near the front cover, and one end of some heat sinks of the heat sink group Ⅱ extends and abuts against the heat generating surface of at least one thermoelectric cooler; A plurality of temperature sensors, which are respectively arranged on the inner side wall of the wind guiding hopper, the outer side wall of the wind guiding hopper, on the substrate near the heat dissipation component and on the inner side wall of the chassis; a plurality of rotational speed sensors, which are respectively electrically connected to the intake fan, the pump and a plurality of exhaust fans; a current sensor, which is electrically connected to at least one thermoelectric cooler; A control subsystem, which includes a deep reinforcement learning module, which is used to collect the detection values of a plurality of temperature sensors, a plurality of rotational speed sensors and a current sensor in real time and normalize the detection values, and input the data obtained by the normalization process into the deep reinforcement learning network architecture in real time to generate an optimized heat dissipation control instruction; a control module, which is communicatively connected to the intake fan, the pump, at least one thermoelectric cooler and a plurality of exhaust fans, and the control module is used to obtain the heat dissipation control instruction in real time, and respectively regulate the opening, closing or operating speed of the intake fan, the pump and a plurality of exhaust fans according to the heat dissipation control instruction, and synchronously regulate the opening, closing or operating current of at least one thermoelectric cooler according to the heat dissipation control instruction.

[0006] Preferably, the heat sink group Ⅰ and the heat sink group Ⅱ are in a louvered structure or a mesh structure, and the overall thickness of the heat sink group Ⅰ or the heat sink group Ⅱ does not exceed 5 cm.

[0007] Preferably, it further includes: a heat insulation layer, which is arranged between the two side covers of the chassis and the heat sink group Ⅱ.

[0008] Preferably, both the heat dissipation box and the refrigeration box are in a sheet-shaped box structure, and the heat dissipation box and the refrigeration box are connected by a pair of diversion pipes, and the pair of diversion pipes are symmetrically distributed on both sides of the intake fan.

[0009] Preferably, it includes: a plurality of air vents opened on the inner structure of the double-layer structure of the air guide funnel; an annular air vent opened at the narrow end of the air guide funnel; a plurality of air guide grooves provided on the inner side wall of the outer structure of the double-layer structure of the air guide funnel, and the plurality of air guide grooves extend along the axial direction of the air guide funnel.

[0010] Preferably, the wide end and the narrow end of the air guide funnel are of circular or rectangular structure.

[0011] Preferably, it includes: a plurality of auxiliary refrigeration chips evenly spaced and arranged on the inner side wall of the air guide funnel, and in the axial direction of the air guide funnel, the plurality of auxiliary refrigeration chips are spirally distributed, and the spacing distance between adjacent two auxiliary refrigeration chips is greater than the aperture of any air vent.

[0012] Preferably, it further includes: two humidity sensors respectively arranged on the outer side wall and inside the double-layer structure of the air guide funnel; a humidity monitoring module for real-time obtaining the relative humidity values of the two humidity sensors and calculating the difference between the two relative humidity values at the same time; if three consecutive differences are all less than or equal to 2% relative humidity, continue to monitor; if three consecutive differences increase and at least one difference is greater than 2% relative humidity, send a warning instruction to the control module, and the control module sends a humidity increase reminder to the CPU of the computer; if three consecutive differences increase and at least two differences are greater than 4% relative humidity, send an alarm instruction to the control module, and the control module sends a water leakage alarm of the water-cooled pipe to the CPU of the computer.

[0013] Preferably, the deep reinforcement learning network architecture is a deep Q network and its derivative algorithms.

[0014] The present invention has at least the following beneficial effects: The air-cooling subsystem is composed of an intake fan and a plurality of exhaust fans arranged on the computer case, and by means of forced air flow, it assists in taking away the heat inside the case; After the water in the water-cooling subsystem is refrigerated by the refrigeration box, the refrigerated water is transported to the water-cooled pipe by a pump. The water-cooled pipe contacts the double-layer structure of the air guide funnel for refrigeration. When the intake fan blows air into the air guide funnel, it will contact the side wall of the air guide funnel and cool down. Then the cooled air is continuously blown to the heating components of the host to perform rapid cooling by combining air cooling and water cooling; the water after heat exchange flows back into the heat dissipation box for preliminary heat dissipation and then flows back to the refrigeration box, which can effectively reduce the temperature difference of the water in the heat dissipation box and the refrigeration box and reduce the refrigeration energy consumption; among them, the water-cooled pipe is arranged in the double-layer structure air guide funnel. On the one hand, it avoids the water-cooled pipe directly contacting the heating components and eliminates the hidden danger of water leakage; on the other hand, the setting structure of the water-cooled pipe is relatively simple and easy to maintain. At the same time, it can effectively increase the refrigeration area and improve the cooling efficiency of air cooling; the application of the semiconductor refrigeration chip has high refrigeration efficiency, and the condensed water generated by the semiconductor refrigeration chip will not affect the humidity inside the case; The heat sink group Ⅰ is used to quickly dissipate the heat in the heat dissipation box, reduce the water temperature, and reduce the refrigeration energy consumption; The heat sink group Ⅱ is used to conduct and dissipate the heat on the heating surface of the semiconductor refrigeration chip in real time to both sides of the chassis, reducing the influence on the air temperature of the intake fan; A plurality of temperature sensors and a plurality of rotational speed sensors are equipped, and a current sensor for detecting the state of at least one semiconductor refrigeration chip is also provided, so as to collect the temperature, power consumption data of computer hardware and the relevant detection values of the semiconductor refrigeration chip in real time, and cooperate effectively with the control subsystem to obtain a good heat dissipation effect with low power consumption; In summary, the computer heat dissipation system based on deep reinforcement learning provided by the present invention combines the air-cooling subsystem, the water-cooling subsystem and the semiconductor refrigeration chip. The deep reinforcement learning module conducts in-depth analysis and learning on the collected and normalized data, continuously interacts with the system environment, continuously optimizes the heat dissipation strategy, and outputs the most suitable heat dissipation control instruction for the control module, so as to ensure that the computer heat dissipation system is always in an efficient and intelligent operating state.

[0015] Other advantages, objectives and features of the present invention will be partially reflected by the following description, and partially will also be understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic structural diagram of the computer heat dissipation system based on deep reinforcement learning described in an embodiment of the present invention; Figure 2 It is a perspective structural diagram of the front cover of the chassis described in an embodiment of the present invention; Figure 3 It is a top view structural diagram of the air-cooling subsystem in the chassis described in an embodiment of the present invention; Figure 4 It is a schematic structural diagram of the water-cooling subsystem described in an embodiment of the present invention; Figure 5 It is a perspective structural diagram of the front cover of the chassis described in another embodiment of the present invention; Figure 6 It is a schematic structural diagram of the cross-section of the air guide funnel described in an embodiment of the present invention; Figure 7 It is a schematic structural diagram of the cross-section of the air guide funnel described in another embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following further elaborates on the present invention with reference to the accompanying drawings, so that those skilled in the art can implement it according to the description in the specification.

[0018] It should be understood that terms such as "having", "including", and "comprising" used herein do not preclude the presence or addition of one or more other elements or combinations thereof.

[0019] As Figures 1 to 4 shown, the present invention provides a computer cooling system based on deep reinforcement learning, including: An air-cooling subsystem 2, which includes an intake fan 201, which is arranged in the middle of the front cover of the computer case 1; a plurality of outlet fans 202, which are respectively arranged on the side cover and the rear cover of the case; a plurality of ventilation holes, which are evenly distributed on the case and are arranged close to the intake fan and the plurality of outlet fans; A water-cooling subsystem 3, which includes a heat dissipation tank 301 and a refrigeration tank 302 that are connected end to end. The heat dissipation tank and the refrigeration tank are both attached to the inner side wall of the front cover of the case, and the heat dissipation tank and the refrigeration tank are respectively located above and below the intake fan; a wind guide hopper 303, which is a funnel-shaped cylinder with a double-layer structure. The wide end of the wind guide hopper faces the intake fan and the refrigeration tank, and the narrow end of the wind guide hopper extends to a position close to the heat-generating components of the computer; a water-cooling pipe 304, one end of which is connected to the liquid outlet end of the refrigeration tank. The main part of the water-cooling pipe is wound inside the double-layer structure of the wind guide hopper, and the other end of the water-cooling pipe is connected to the water inlet end of the heat dissipation tank; a pump 305, which is arranged on the water-cooling pipe; at least one semiconductor refrigeration sheet 306, whose refrigerating surface is attached to the side wall of the refrigeration tank, and the heating surface of the semiconductor refrigeration sheet faces the outside of the case through the air vent on the front cover; a heat sink group Ⅰ 307, which is arranged on the front cover of the case, and the heat sink group Ⅰ abuts against the outside of the heat dissipation tank; a heat sink group Ⅱ 308, which is arranged on both sides of the case close to the front cover, and one end of some heat sinks of the heat sink group Ⅱ extends and abuts against the heating surface of at least one semiconductor refrigeration sheet; a plurality of temperature sensors, which are respectively arranged on the inner side wall of the wind guide hopper, the outer side wall of the wind guide hopper, the substrate close to the heat dissipation component, and the inner side wall of the case; a plurality of rotational speed sensors, which are respectively electrically connected to the intake fan, the pump, and the plurality of outlet fans; a current sensor, which is electrically connected to at least one semiconductor refrigeration sheet; A control subsystem 4, which includes a deep reinforcement learning module, which is used to collect the detection values of a plurality of temperature sensors, a plurality of rotational speed sensors, and a current sensor in real time and normalize the detection values, and input the data obtained by the normalization process into the deep reinforcement learning network architecture in real time to generate an optimized heat dissipation control instruction; a control module, which is communicatively connected to the intake fan, the pump, at least one semiconductor refrigeration sheet, and the plurality of outlet fans. The control module is used to obtain the heat dissipation control instruction in real time, and respectively regulate the opening, closing, or operating speed of the intake fan, the pump, and the plurality of outlet fans according to the heat dissipation control instruction, and synchronously regulate the opening, closing, or operating current of at least one semiconductor refrigeration sheet according to the heat dissipation control instruction.

[0020] In this solution, the air-cooling subsystem consists of an intake fan and multiple exhaust fans arranged on the computer chassis. By means of forced air flow, it helps to remove the heat inside the chassis. After the water in the water-cooling subsystem is cooled by the refrigeration box, the cooled water is pumped to the water-cooling pipe by a pump. The water-cooling pipe contacts the double-layer structure of the air guide funnel for refrigeration. When the intake fan blows air into the air guide funnel, it will contact the side wall of the air guide funnel and cool down. Then the cooled air is continuously blown onto the heat-generating components of the host computer for rapid cooling by combining air cooling and water cooling. The water after heat exchange flows back into the heat dissipation box for preliminary heat dissipation and then returns to the refrigeration box, which can effectively reduce the temperature difference of the water in the heat dissipation box and the refrigeration box and reduce the refrigeration energy consumption. Among them, the water-cooling pipe is arranged in the double-layer air guide funnel. On the one hand, it avoids the direct contact between the water-cooling pipe and the heat-generating components, eliminating the hidden danger of water leakage. On the other hand, the setting structure of the water-cooling pipe is relatively simple and easy to maintain. At the same time, it can effectively increase the refrigeration area and improve the cooling efficiency of air cooling. The application of the thermoelectric cooler has high refrigeration efficiency, and the condensed water generated by the thermoelectric cooler will not affect the humidity inside the chassis. The heat sink group Ⅰ is used to quickly dissipate the heat in the heat dissipation box, reduce the water temperature, and reduce the refrigeration energy consumption. The heat sink group Ⅱ is used to conduct and dissipate the heat on the heating surface of the thermoelectric cooler to both sides of the chassis in real time, reducing the influence on the air temperature of the intake fan. Multiple temperature sensors and multiple rotational speed sensors are equipped, and a current sensor for detecting the state of at least one thermoelectric cooler is also provided. In this way, the temperature and power consumption data of the computer hardware, as well as the relevant detection values of the thermoelectric cooler, are collected in real time, and they cooperate effectively with the control subsystem to obtain a good heat dissipation effect with low power consumption. In summary, the computer heat dissipation system based on deep reinforcement learning provided by the present invention combines the air-cooling subsystem, the water-cooling subsystem and the thermoelectric cooler. The deep reinforcement learning module conducts in-depth analysis and learning on the collected and normalized data, continuously interacts with the system environment, and continuously optimizes the heat dissipation strategy, outputting the most suitable heat dissipation control instruction for the control module, so as to ensure that the computer heat dissipation system is always in an efficient and intelligent operating state.

[0021] In a preferred solution, the heat sink group Ⅰ and the heat sink group Ⅱ are in a louver structure or a mesh structure, and the overall thickness of the heat sink group Ⅰ or the heat sink group Ⅱ does not exceed 5 cm.

[0022] As Figure 5 shown, in a preferred solution, it further includes: a heat insulation layer 309, which is arranged between the two side covers of the chassis and the heat sink group Ⅱ. So as to isolate the heat on the heat sink Ⅱ outside the chassis and reduce its influence on the temperature inside the chassis.

[0023] In a preferred embodiment, both the heat dissipation box and the refrigeration box are sheet-shaped box structures. The heat dissipation box and the refrigeration box are connected through a pair of diversion pipes, and the pair of diversion pipes are symmetrically distributed on both sides of the intake fan.

[0024] As Figure 6 shown, in a preferred embodiment, it includes: a plurality of ventilation holes 310, which are opened on the inner structure 3011 of the double-layer structure of the air guide funnel; an annular ventilation opening, which is opened at the narrow end of the air guide funnel; a plurality of air guide grooves 311, which are arranged on the inner side wall of the outer structure 3012 of the double-layer structure of the air guide funnel, and the plurality of air guide grooves extend along the axial direction of the air guide funnel. Most of the air poured in by the intake fan is directly diverted to the heating component after being cooled by the air guide funnel; a small part is blown into the double-layer structure through the plurality of ventilation holes, makes more sufficient contact with the water-cooled pipe for refrigeration, and then is blown to the heating component through the plurality of air guide grooves and the annular ventilation opening, further improving the heat dissipation efficiency.

[0025] In a preferred embodiment, the wide end and the narrow end of the air guide funnel are circular or rectangular structures. According to the structural requirements, the wide end of the air guide funnel can be set to be circular to better fit the structure of the intake fan, while the narrow end of the air guide funnel can be set to be rectangular to fit the structure of the heating component.

[0026] As Figure 7 shown, in a preferred embodiment, it includes: a plurality of auxiliary refrigeration sheets 312, which are evenly spaced and arranged on the inner side wall of the air guide funnel, and in the axial direction of the air guide funnel, the plurality of auxiliary refrigeration sheets are spirally distributed, and the distance between adjacent two auxiliary refrigeration sheets is greater than the aperture of any ventilation hole. The plurality of auxiliary refrigeration sheets can increase the refrigeration area of the air guide funnel. When the wind blows through the air guide funnel, it effectively improves the refrigeration effect of the wind, and the plurality of auxiliary refrigeration sheets distributed in a spiral shape also divert the wind to a certain extent, increasing the refrigeration time of the wind and further improving the refrigeration effect.

[0027] In a preferred embodiment, it further includes: two humidity sensors, which are respectively arranged on the outer side wall of the air guide funnel and inside the double-layer structure; a humidity monitoring module, which is used to obtain the relative humidity values of the two humidity sensors in real time and calculate the difference between the two relative humidity values at the same time; if the three consecutive differences are all less than or equal to 2% relative humidity, continue to monitor; if the three consecutive differences increase and at least one difference is greater than 2% relative humidity, send a warning instruction to the control module, and the control module sends a humidity increase reminder to the CPU of the computer; if the three consecutive differences increase and at least two differences are greater than 4% relative humidity, send an alarm instruction to the control module, and the control module sends a water-cooled pipe leakage alarm to the CPU of the computer.

[0028] In a preferred embodiment, the deep reinforcement learning network architecture is constructed based on the deep Q-network and its derivative algorithms. The deep Q-network (DQN) combines deep learning and Q-learning, and approximates the Q-function by constructing a deep neural network. This network takes the system state as input and outputs the Q-values corresponding to each possible action. During training, the Bellman equation is used to calculate the target Q-value, and the network parameters are updated by minimizing the mean square error between the predicted Q-value and the target Q-value. For example, in a computer cooling system, state information such as CPU temperature and load is input into the DQN network, and the network outputs the Q-values of different actions such as increasing the rotation speed of the intake fan, increasing the number of activated exhaust fans, increasing the rotation speed of the existing exhaust fans, or increasing the rotation speed of the pump. The model selects the action with the largest Q-value to execute.

[0029] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the examples shown and described herein.

Claims

1. A computer cooling system based on deep reinforcement learning, characterized in that, Comprising: An air-cooling subsystem, which includes an intake fan, which is arranged in the middle of the front cover of the computer case; Multiple exhaust fans, which are respectively arranged on the side cover and the rear cover of the case; multiple ventilation holes, which are evenly distributed on the case and are arranged close to the intake fan and the multiple exhaust fans; A water-cooling subsystem, which includes a heat dissipation box and a refrigeration box that are connected end to end. The heat dissipation box and the refrigeration box are both attached and arranged on the inner side wall of the front cover of the case, and the heat dissipation box and the refrigeration box are respectively located above and below the intake fan; a wind guide funnel, which is a double-layered funnel-shaped cylinder. The wide end of the wind guide funnel faces the intake fan and the refrigeration box, and the narrow end of the wind guide funnel extends to a position close to the heat-generating components of the computer; a water-cooling pipe, one end of which is connected to the liquid outlet end of the refrigeration box, the main part of the water-cooling pipe is coiled inside the double-layer structure of the wind guide funnel, and the other end of the water-cooling pipe is connected to the water inlet end of the heat dissipation box; A pump, which is arranged on the water-cooling pipe; at least one semiconductor refrigeration chip, the refrigerating surface of which is attached to the side wall of the refrigeration box, and the heating surface of the semiconductor refrigeration chip faces the outside of the case through the air vent on the front cover; a heat sink group Ⅰ, which is arranged on the front cover of the case, and the heat sink group Ⅰ abuts against the outside of the heat dissipation box; A heat sink group Ⅱ, which is arranged on both sides of the case close to the front cover, and one end of some heat sinks of the heat sink group Ⅱ extends and abuts against the heating surface of at least one semiconductor refrigeration chip; Multiple temperature sensors, which are respectively arranged on the inner side wall of the wind guide funnel, the outer side wall of the wind guide funnel, the substrate close to the heat dissipation component, and the inner side wall of the case; multiple rotational speed sensors, which are respectively electrically connected to the intake fan, the pump, and the multiple exhaust fans; a current sensor, which is electrically connected to at least one semiconductor refrigeration chip; A control subsystem, which includes a deep reinforcement learning module, which is used to collect the detection values of the multiple temperature sensors, the multiple rotational speed sensors, and the current sensor in real time and normalize the detection values, and input the data obtained by the normalization process into the deep reinforcement learning network architecture in real time to generate an optimized heat dissipation control instruction; a control module, which is communicatively connected to the intake fan, the pump, at least one semiconductor refrigeration chip, and the multiple exhaust fans. The control module is used to obtain the heat dissipation control instruction in real time, and respectively regulate the opening, closing, or operating rotational speed of the intake fan, the pump, and the multiple exhaust fans according to the heat dissipation control instruction, and synchronously regulate the opening, closing, or operating current of at least one semiconductor refrigeration chip according to the heat dissipation control instruction.

2. The computer cooling system based on deep reinforcement learning according to claim 1, characterized in that, The heat sink group Ⅰ and the heat sink group Ⅱ are in a louver structure or a mesh structure, and the overall thickness of the heat sink group Ⅰ or the heat sink group Ⅱ does not exceed 5 cm.

3. The computer cooling system based on deep reinforcement learning according to claim 1, characterized in that, It also includes: A heat insulation layer, which is arranged between the two side covers of the case and the heat sink group Ⅱ.

4. The computer cooling system based on deep reinforcement learning according to claim 1, characterized in that, Both the heat dissipation box and the refrigeration box are in a sheet-shaped box structure, and the heat dissipation box and the refrigeration box are connected by a pair of diversion pipes, and the pair of diversion pipes are symmetrically distributed on both sides of the intake fan.

5. The computer cooling system based on deep reinforcement learning according to claim 1, characterized in that, Comprising: Multiple air vents, which are opened on the inner layer structure of the double-layer structure of the wind guide funnel; An annular air vent, which is opened on the narrow end of the wind guide funnel; Multiple wind guide grooves, which are arranged on the inner side wall of the outer layer structure of the double-layer structure of the wind guide funnel, and the multiple wind guide grooves extend along the axial direction of the wind guide funnel.

6. The computer cooling system based on deep reinforcement learning according to claim 1, wherein, The wide end and the narrow end of the air guide funnel are of circular or rectangular structure.

7. The computer cooling system based on deep reinforcement learning according to claim 5, characterized in that, It includes: A plurality of auxiliary cooling fins, which are evenly spaced on the inner side wall of the air guide funnel, and in the axial direction of the air guide funnel, the plurality of auxiliary cooling fins are spirally distributed, and the spacing distance between two adjacent auxiliary cooling fins is greater than the aperture of any air vent.

8. The computer heat dissipation system based on deep reinforcement learning according to claim 1, characterized in that, It further includes: Two humidity sensors, which are respectively arranged on the outer side wall of the air guide funnel and inside the double-layer structure; A humidity monitoring module, which is used to obtain the relative humidity values of the two humidity sensors in real time and calculate the difference between the two relative humidity values at the same time; if three consecutive differences are less than or equal to 2% relative humidity, continue to monitor; if three consecutive differences increase and at least one difference is greater than 2% relative humidity, send a warning instruction to the control module, and the control module sends a humidity increase reminder to the CPU of the computer; if three consecutive differences increase and at least two differences are greater than 4% relative humidity, send an alarm instruction to the control module, and the control module sends a water-cooled pipe leakage alarm to the CPU of the computer.

9. The computer cooling system based on deep reinforcement learning according to claim 1, characterized in that, The deep reinforcement learning network architecture is the deep Q network and its derivative algorithms.

Images