Refrigerator

Abstract

The invention discloses a refrigerator which comprises an evaporator, a fan, a defrosting element and a processing device. The processing device comprises a storage part, an acquisition part, a selection part and a control part. A prediction model constructed based on a deep Q network is stored in the state space, the state space comprises operation parameters of the refrigerator, and the action space comprises a plurality of selectable defrosting durations; the defrosting duration prediction module is configured to obtain operation parameters of the refrigerator in the current working period and input the operation parameters into the prediction model, and the selectable defrosting duration with the maximum Q value is selected from output of the prediction model to serve as predicted defrosting duration; when the running duration of the defrosting element reaches the predicted defrosting duration, the defrosting element is closed; the Q value represents an expected value for executing one selectable defrosting duration under the operation parameters of the refrigerator in the current working period. By introducing the prediction model constructed based on the deep Q network, the defrosting process of the evaporator can be intelligently optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of refrigeration equipment, and in particular to a refrigerator. Background Art

[0002] Air-cooled refrigerators are devices that use fans to circulate cold air to cool the interior space. Compared with direct-cooled refrigerators, air-cooled refrigerators have a more even temperature distribution in the storage compartment, and frost will not form in the storage compartment. Users do not need to manually defrost regularly, and the user experience is better.

[0003] However, although frost will not form in the storage room, after a period of operation, the humid air in each storage room of the refrigerator will condense on the evaporator to form a frost layer. When the frost layer accumulates to a certain thickness, the heat exchange efficiency of the evaporator will be significantly reduced, increasing the overall energy consumption of the air-cooled refrigerator. When the thickness of the frost layer accumulates to a certain extent, it may also cause the cold air to be unable to circulate, causing the temperature in the storage room to gradually rise and lose the cooling capacity.

[0004] To solve this problem, an electric defrost element is usually set on the evaporator in the prior art. After the refrigerator has been working for a certain period of time, the refrigeration is stopped and the electric defrost element is started to defrost. However, there is no economical and effective detection means in the prior art to determine whether the frost layer on the evaporator is completely melted. A more common practice is to set a temperature sensor on the upper pipe of the evaporator to indirectly determine whether the frost layer on the evaporator is completely melted by the detection value of the temperature sensor. Specifically, a defrost exit temperature (for example, 5 degrees Celsius) is set, and when the temperature detection value reaches the defrost exit temperature, it is considered that the frost layer is completely melted. The operation of the electric defrost element will cause additional energy consumption and increase the compartment temperature, which will increase the power consumption of the air-cooled refrigerator. But in fact, due to the uneven frost on the evaporator, the melting speed of the frost layer in each part is also different. In order to ensure that the frost layer at each position is completely melted, the defrost exit temperature will be set relatively high, which will increase the power consumption of the refrigerator.

[0005] The above information disclosed in the background technology is only used to increase the understanding of the background technology of the present application, and therefore, it may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] In view of the problem that the setting value of the defrost exit temperature in the prior art is relatively high, which will additionally increase the power consumption of the refrigerator, the first aspect of the present application provides a refrigerator.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions:

[0008] In one or more embodiments of the present application, a refrigerator is provided, comprising an evaporator, a fan and a defrost element; wherein the fan is used to guide air heat-exchanged with the evaporator into a storage compartment; and the defrost element is used to melt the frost layer on the evaporator.

[0009] In one or more embodiments of the present application, the refrigerator also includes a processing device, the processing device includes: a storage unit, an acquisition unit, a selection unit and a control unit; wherein the storage unit stores a prediction model constructed based on a deep Q network, the state space of the prediction model includes the operating parameters of the refrigerator, and the action space includes multiple optional defrost time lengths; the acquisition unit is used to obtain the operating parameters of the refrigerator in the current working cycle; the selection unit is configured to input the operating parameters of the refrigerator in the current working cycle into the prediction model, and select the optional defrost time length with the largest Q value in the output of the prediction model as the predicted defrost time length; the control unit is configured to turn off the defrost element when the operating time of the defrost element reaches the predicted defrost time length.

[0010] In one or more embodiments of the present application, the Q value represents an expected value of executing one of the optional defrosting durations under the operating parameters of the refrigerator in the current working cycle.

[0011] In one or more embodiments of the present application, the processing device further includes: a training module, which is used to train the prediction model.

[0012] In one or more embodiments of the present application, the training module includes a setting unit, an execution unit and a generation unit; the setting unit is configured to set a prediction network and a target network; the execution unit is configured to obtain a first operating parameter of a refrigerator with a first working cycle and input it into the prediction network, and adopt an ε-greedy strategy to select an optional defrost time as the defrost time to be evaluated; drive the defrost element to operate, and when the operating time of the defrost element reaches the defrost time to be evaluated, turn off the defrost element; the generation unit is configured to obtain a second operating parameter of a refrigerator with a second working cycle and input it into the prediction network, select the maximum Q value from the output of the prediction network, and generate a reward value to obtain sample data.

[0013] In one or more embodiments of the present application, in a training state, the execution unit is configured to determine whether the working time of the refrigerator exceeds a preset frosting cycle time; when the working time of the refrigerator exceeds the frosting cycle time, obtain the first operating parameters of the refrigerator in the first working cycle and input them into the prediction network.

[0014] In one or more embodiments of the present application, in a training state, the generating unit is configured to re-accumulate the working time of the refrigerator after turning off the defrost element, and determine whether the re-accumulated working time of the refrigerator exceeds the preset evaluation cycle time; when the re-accumulated working time of the refrigerator exceeds the preset evaluation cycle time, obtain the second operating parameters of the refrigerator in the second working cycle and input them into the prediction network.

[0015] In one or more embodiments of the present application, the duration of the frosting cycle is greater than the duration of the evaluation cycle.

[0016] In one or more embodiments of the present application, the training module further includes: a collection unit, a sampling unit, a target Q value calculation unit, a loss value calculation unit and an update unit.

[0017] In one or more embodiments of the present application, the collection unit is configured to collect sample data and store the sample data in an experience pool D, each sample data including: a first operating parameter of a refrigerator with a first working cycle, a defrost time to be evaluated, a reward value, and a second operating parameter of a refrigerator with a second working cycle.

[0018] In one or more embodiments of the present application, the sampling unit is configured to infer whether the number of sample data in the experience pool is higher than a preset number threshold; if the number of sample data in the experience pool is higher than the preset number threshold, batch sample data corresponding to the preset number threshold is extracted from the experience pool.

[0019] In one or more embodiments of the present application, the target Q value calculation unit is configured to input the first operating parameter of the first working cycle refrigerator in the batch sample data into the prediction network to obtain the first maximum Q value output by the prediction network; input the second operating parameter of the second working cycle refrigerator in the batch sample data into the target network to obtain the second maximum Q value output by the target network, and calculate the target Q value based on the second maximum Q value, the reward value in the batch sample data, and the discount factor.

[0020] In one or more embodiments of the present application, the loss value calculation unit is configured to calculate the loss values ​​of the target network and the prediction network according to the first maximum Q value and the target Q value.

[0021] In one or more embodiments of the present application, the updating unit is configured to update the configuration parameters of the prediction network by back propagation.

[0022] In one or more embodiments of the present application, the training module further includes a counting unit and a judging unit. The counting unit is configured to increase the count value of the training counter by 1 after the updating unit updates the configuration parameters of the prediction network; the judging unit is configured to update the configuration parameters of the target network when the count value of the training counter is divisible by the preset freezing round.

[0023] In one or more embodiments of the present application, the training module further includes an exit unit. The exit unit is configured to exit training when the loss value is lower than a preset training error or the count value of the training counter exceeds a preset total training round; and use the prediction network as the prediction model.

[0024] In one or more embodiments of the present application, the reward value is generated based on a difference between a first operating parameter of the first working cycle refrigerator and a second operating parameter of the second working cycle refrigerator, and based on energy consumption corresponding to the defrost time to be evaluated.

[0025] In one or more embodiments of the present application, the operating parameter of the working cycle is an average value of the operating parameter sampled within the working cycle.

[0026] In one or more embodiments of the present application, the operating parameters of the working cycle include one or more of indoor ambient temperature, storage compartment set temperature, storage compartment actual temperature, evaporator surface temperature, compressor frequency and compressor power.

[0027] Compared with the prior art, the advantages and positive effects of the present invention are: the refrigerator provided by the present application can intelligently optimize the defrosting process of the evaporator by introducing a prediction model based on a deep Q network. Not only can the defrosting time be dynamically adjusted according to the actual operating conditions to improve the defrosting efficiency, but also unnecessary energy consumption can be reduced, thereby achieving more intelligent and efficient refrigerator management.

[0028] After reading the specific embodiments of the present invention in conjunction with the accompanying drawings, other features and advantages of the present invention will become more clear. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0030] Figure 1 A schematic structural block diagram of a refrigerator provided by some embodiments of the present invention;

[0031] Figure 2 A schematic diagram of the structure of a refrigerator provided by some embodiments of the present invention;

[0032] Figure 3 A schematic structural block diagram of a refrigerator provided by some embodiments of the present invention;

[0033] Figure 4A schematic block diagram of the structure of a processing device in a refrigerator provided by some embodiments of the present invention;

[0034] Figure 5 A flow chart of a processing device in a refrigerator provided by some embodiments of the present invention;

[0035] Figure 6 A schematic block diagram of the structure of a processing device in a refrigerator provided by some embodiments of the present invention;

[0036] Figure 7 A schematic block diagram of the structure of a training module in a refrigerator provided in some embodiments of the present invention;

[0037] Figure 8 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0038] Fig. 9 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0039] Fig.10 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0040] Fig.11 A schematic block diagram of the structure of a training module in a refrigerator provided in some embodiments of the present invention;

[0041] Fig.12 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0042] Fig.13 A schematic block diagram of the structure of a training module in a refrigerator provided in some embodiments of the present invention;

[0043] Fig.14 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0044] Fig.15 A schematic block diagram of the structure of a training module in a refrigerator provided in some embodiments of the present invention;

[0045] Fig.16 A schematic block diagram of the structure of a training module in a refrigerator provided in some embodiments of the present invention;

[0046] Fig.17 A flowchart of a training module in a refrigerator provided by some embodiments of the present invention;

[0047] In the figure: 1. refrigerator; 10. refrigerator body;

[0048] 11. Door; 12. Drawer; 13. Controller; 14. Server; 15. Terminal device; 161. Compressor; 162. Fan; 163. Defrost heater;

[0049] 20. Processing device; 201. Storage unit; 202. Acquisition unit; 203. Selection unit; 204. Control unit;

[0050] 30. Sensor assembly; 31. Compartment temperature sensor; 32. Evaporator temperature sensor; 33. Ambient temperature sensor; 34. Electrical parameter sensor;

[0051] 40. Training module; 401. Setting unit; 402. Execution unit; 403. Generation unit; 404. Collection unit; 405. Sampling unit; 406. Target Q value calculation unit; 407. Loss value calculation unit; 408. Update unit; 409. Counting unit; 410. Judgment unit; 411. Exit unit. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0054] The terms "first", "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0055] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0056] In the present invention, unless otherwise clearly stipulated and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0057] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplicity and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed. In addition, the present invention provides various specific examples of processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and / or the use of other materials.

[0058] When judging whether the frost layer on the evaporator is completely melted by the detection value of the temperature sensor, due to the uneven frost on the evaporator, the melting speed of the frost layer in different parts is also different. In order to ensure that the frost layer in each position is completely melted, a higher defrost exit temperature is set, which increases the power consumption of the refrigerator. One or more methods of the present application design and provide a refrigerator.

[0059] Hereinafter, the refrigerator 1 provided by the present application will be described with reference to the accompanying drawings. The overall structure of the refrigerator 1 is as follows: Figure 1 The refrigerator 1 includes a refrigerator body 10 and a processing device 20 .

[0060] like Figure 2 As shown, a storage compartment is constructed in the refrigerator body 10 .

[0061] Storage compartments are usually divided into freezer compartments and refrigerated compartments, and can also be divided into compartments with special functions, such as compartments for storing fruits and vegetables. The refrigerated compartment can be kept at a temperature range of about 4°C to store food, medicine or biological agents in a refrigerated state. The freezer compartment can be kept at a temperature range of about -18°C to store food, medicine or biological agents in a frozen state.

[0062] The storage compartment has an opening, which can be opened and closed by a door 11 hinged to the housing, or opened and closed by a drawer 12. When a refrigerating compartment and a freezing compartment are provided, one of the openings can be opened and closed by the door (e.g., the refrigerating compartment), and the other opening can be opened and closed by the drawer 12 (e.g., the freezing compartment).

[0063] The refrigerator body 10 uses a vapor compression refrigeration cycle to generate energy for maintaining the target temperature. The refrigeration cycle is composed of a compressor 161, a condenser, a throttling device and an evaporator. The refrigeration cycle involves a series of processes, including compression, condensation, expansion and evaporation, to cool the storage compartment to maintain an ideal low-temperature storage environment inside.

[0064] In the vapor compression refrigeration cycle, low-temperature and low-pressure refrigerant enters the compressor 161, which compresses the refrigerant into high-temperature and high-pressure refrigerant gas and discharges the compressed refrigerant gas. The discharged refrigerant gas flows into the condenser, which condenses the compressed refrigerant into a liquid phase, and heat is released to the surrounding environment through the condensation process.

[0065] The throttling device expands the liquid refrigerant in a high temperature and high pressure state formed in the condenser into a low pressure liquid refrigerant. The evaporator evaporates the refrigerant expanded in the throttling device and returns the refrigerant gas in a low temperature and low pressure state to the compressor 161. The evaporator can achieve a refrigeration effect by heat exchange with the material to be cooled through the latent heat of evaporation of the refrigerant. In the present application, the evaporator performs heat exchange with the air to form air for cooling the storage compartment, thereby achieving cooling of the storage compartment. The throttling device can be a capillary tube.

[0066] A filter is also provided downstream of the condenser to filter out impurities in the refrigerant, thereby improving the heat exchange efficiency of the refrigeration unit and reducing the risk of pipeline blockage.

[0067] A liquid storage tank may also be provided on the suction side of the compressor 161. The liquid storage tank is used to separate the refrigerant from gas and liquid. The liquid storage tank is a shell-shaped component. The gas-liquid mixed refrigerant fluid enters the liquid storage tank for basic phase separation. The gas enters the gas channel for gravity sedimentation to separate the droplets. The liquid enters the liquid space to separate the bubbles. The gas flows out from the gas outlet and is sucked into the compressor 161, so as to avoid the compressor 161 from sucking in liquid and reducing the service life of the compressor 161.

[0068] The compressor 161 and the condenser can be arranged on the rear lower side of the shell, and the evaporator can be arranged at the position corresponding to the storage compartment on the rear side of the shell. The evaporator and the condenser can also be arranged at other positions according to the industrial design of the refrigerator body 10, which are not listed here one by one. The position where the evaporator is arranged has a certain space, allowing air to flow therein. The air is driven by the fan 162 to guide the air formed by the evaporator for cooling the storage compartment into the storage compartment, and to inhale the air in the storage compartment to form an air circulation. In one or more embodiments of the present application, the fan 162 includes a refrigeration fan and a freezing fan. In one or more embodiments of the present application, the fan 162 can also be arranged in conjunction with the condenser.

[0069] In one or more embodiments of the present application, the evaporator may also be divided into two parts for the refrigerating compartment and the freezing compartment, which are referred to as a refrigerating compartment cooler and a freezing compartment cooler.

[0070] A defrosting element is provided in the refrigerator body 10, and the defrosting element is used to melt the frost layer on the evaporator.

[0071] In one or more embodiments of the present application, the defrost element includes a defrost heater 163 , which may be an electric heating tape or an electric heater.

[0072] In one or more embodiments of the present application, the defrost element may also be a combination of an electric heating tape or an electric heater, and a heat exchanger or a heat exchange pipeline. When the defrosting conditions are met, the heat exchange pipeline is turned on by controlling the combination of the solenoid valve or the valve element, and the high-temperature and high-pressure refrigerant discharged from the compressor 161 enters the heat exchange pipeline, exchanges heat with the surrounding air, increases the temperature of the air, and further provides heat to melt the frost layer on the surface of the evaporator so that the evaporator is in a defrosted state. The heat exchange pipeline can be arranged below the evaporator, and the air is guided to remove the ice or frost layer on the evaporator by using the principle that the hot air flow has a low density and will rise. The defrost element composed of the electric heating tape or the electric heater can also be arranged at other positions around the evaporator, such as on the upper side or one side of the evaporator.

[0073] In one or more embodiments of the present application, the processing device is a system on a chip including a processor, a memory, and a communication interface.

[0074] In one or more embodiments of the present application, the processing device 20 may be communicatively connected with the controller 13 , for example, implemented by a terminal device 15 and / or a cloud server 14 .

[0075] In one or more embodiments of the present application, part of the functions of the processing device 20 may be implemented by the controller 13 , and part of the functions may be implemented by the terminal device 15 and / or the cloud server 14 .

[0076] The controller 13 can communicate with the terminal device 15 and / or the server 14. The network between the controller 13 and the terminal device 15, or the controller 13 and the server 14 can be the Internet, a cellular network, a Wi-Fi network, a low power wide area network (Low Power Wide Area) based on standards and protocols such as LoRa, Sigfox, NB-IoT, a wide area network, a local area network, and the like.

[0077] The refrigerator body 10 can be used in a home environment to store daily necessities such as food and cold drinks; it can also be used in commercial places such as restaurants, hotels, supermarkets, convenience stores, etc. to store ingredients, food and beverages to meet customer needs; it can also be used in places such as hospitals and laboratories to store medicines and biological samples to meet medical and scientific research needs.

[0078] The server 14 can provide various network services, such as providing resources and data access for the refrigerator 1 controller 13 and the terminal device 15. The server 14 has higher performance and reliability. The server 14 can be connected to the controllers 13 of multiple refrigerators 1, multiple terminal devices 15, and other smart home appliance terminals.

[0079] The terminal device 15 is an electronic device with intelligent functions. It can be connected to the above-mentioned various networks to realize functions such as remote control, data exchange and human-computer interaction. The terminal device 15 includes smart phones, tablet computers, smart speakers, wearable devices, smart home appliances (such as smart TVs) and smart car devices, etc. The interaction between the terminal device 15 and the user includes but is not limited to: operating on the screen with fingers or touch pens, realizing various operations through buttons, voice control, gesture control, iris recognition, and facial recognition, etc.

[0080] In one or more embodiments of the present application, the refrigerator body 10 is in communication connection with the sensor assembly 30. At least part of the sensors in the sensor assembly 30 are disposed in the refrigerator body 10.

[0081] like Figure 3 As shown, in one or more embodiments of the present application, the sensor assembly 30 includes a compartment temperature sensor 31 , an evaporator temperature sensor 32 , and an ambient temperature sensor 33 .

[0082] Exemplarily, the compartment temperature sensor 31 includes a refrigerating compartment temperature sensor and a freezing compartment temperature sensor. The refrigerating compartment temperature sensor is disposed in the refrigerating compartment of the refrigerator body to detect the temperature of the refrigerating compartment; the freezing compartment temperature sensor is disposed in the freezing compartment of the refrigerator body to detect the temperature of the freezing compartment.

[0083] In one or more embodiments of the present application, the compartment temperature sensor 31 also includes a fruit and vegetable compartment temperature sensor.

[0084] In one or more embodiments of the present application, the compartment temperature sensor 31 further includes a variable temperature compartment temperature sensor.

[0085] Exemplarily, the evaporator temperature sensor 32 is disposed at the evaporator for detecting the temperature of the evaporator.

[0086] In one or more embodiments of the present application, the evaporator temperature sensor includes a refrigerating compartment cooler temperature sensor and a freezing compartment cooler temperature sensor.

[0087] In one or more embodiments of the present application, the ambient temperature sensor 33 includes an indoor temperature sensor.

[0088] In one or more embodiments of the present application, the ambient temperature sensor 33 includes an indoor temperature sensor and an outdoor temperature sensor (not shown), and the outdoor temperature can also be obtained by querying the server.

[0089] In one or more embodiments of the present application, the sensor assembly 30 further includes a humidity sensor.

[0090] In one or more embodiments of the present application, the sensor assembly 30 further includes a door switch sensor to detect the opening and closing of the door 11 .

[0091] In one or more embodiments of the present application, the sensor assembly 30 may further include other sensors, such as a vibration sensor, a weight sensor, and the like.

[0092] In one or more embodiments of the present application, the sensor assembly 30 further includes an electrical parameter sensor 34. The number of electrical parameter sensors 34 is not limited, and the electrical parameter sensor 34 can be used to detect one or more of power, power peak, power valley, current, voltage, and energy efficiency.

[0093] like Figure 4 As shown, in one or more embodiments of the present application, the processing device 20 includes a storage unit 201, an acquisition unit 202, a selection unit 203, and a control unit 204. Each of these components can be implemented by a processor running a program.

[0094] In one or more embodiments of the present application, the storage unit 201 stores a prediction model constructed based on a deep Q network.

[0095] In one or more embodiments of the present application, the state space of the prediction model includes operating parameters of a refrigerator; the environment of the deep Q network is the refrigerator. The state space defines all possible states of the environment, and a state in the state space is a description of the refrigerator (i.e., the environment) at a certain moment. More specifically, the operating parameters in the state space affect the decision of the prediction model.

[0096] In one or more embodiments of the present application, the operating parameters of the refrigerator include one or more of indoor ambient temperature, storage compartment set temperature, storage compartment actual temperature, evaporator surface temperature, compressor frequency and compressor power.

[0097] In one or more embodiments of the present application, the operating parameters of the refrigerator may further include other operating parameters not listed.

[0098] In one or more embodiments of the present application, the state space can be expressed as: in represents the first operating parameter of the refrigerator in the first working cycle, Indicates the second operating parameter of the refrigerator in the second working cycle.

[0099] In one or more embodiments of the present application, the first operating parameter of the refrigerator in the first working cycle is an average value of the operating parameters sampled in the first working cycle.

[0100] In one or more embodiments of the present application, the second operating parameter of the refrigerator in the second working cycle is an average value of the operating parameters sampled in the second working cycle.

[0101] In one or more embodiments of the present application, the first operating parameter of the refrigerator in the first working cycle is Stored in the form of a two-dimensional tensor.

[0102] In one or more embodiments of the present application, the second operating parameter of the refrigerator in the second working cycle is Stored in the form of a two-dimensional tensor.

[0103] In one or more embodiments of the present application, the action space of the prediction model includes a plurality of optional defrosting durations.

[0104] In one or more embodiments of the present application, the action space can be expressed as:

[0105] A={6,8,10,12,14,16,18,20,22,24,26,28,30}

[0106] The optional defrost time is in minutes.

[0107] In one or more embodiments of the present application, the acquisition unit 202 is used to acquire the operating parameters of the refrigerator in the current working cycle. The operating parameters of the refrigerator in the current working cycle are expressed as

[0108] In one or more embodiments of the present application, the operating parameters of the refrigerator in the current working cycle are It is the average value of the operating parameters sampled in the current working cycle.

[0109] In one or more embodiments of the present application, the operating parameters of the refrigerator in the current working cycle are Stored in the form of a two-dimensional tensor.

[0110] In one or more embodiments of the present application, the selection unit 203 is configured to set the operating parameters of the refrigerator in the current working cycle to The input is input into the prediction model, and the optional defrost time with the largest Q value is selected as the predicted defrost time in the output of the prediction model. The Q value represents the expected value of executing one of the optional defrost times under the operating parameters of the refrigerator in the current working cycle.

[0111] In one or more embodiments of the present application, the control unit 204 is configured to turn off the defrost element when the operating time of the defrost element reaches the predicted defrosting time.

[0112] The refrigerator provided in this application can intelligently optimize the defrosting process of the evaporator by introducing a prediction model based on a deep Q network. It can not only dynamically adjust the defrosting time according to the actual operating conditions and improve the defrosting efficiency, but also reduce unnecessary energy consumption, thereby achieving more intelligent and efficient refrigerator management.

[0113] The processing device 20 is configured to perform the following steps: Figure 5 Multiple steps shown.

[0114] Step S101: obtaining the operating parameters of the refrigerator in the current working cycle.

[0115] Step S102: inputting the operating parameters of the refrigerator in the current working cycle into the prediction model.

[0116] Step S103: Select the optional defrost duration with the largest Q value from the output of the prediction model as the predicted defrost duration.

[0117] Step S104: determining whether the operation time of the defrost element reaches the predicted defrost time.

[0118] Step S105: If the operation time of the defrost element reaches the predicted defrost time, the defrost element is turned off.

[0119] Step S106: If the operation time of the defrost element does not reach the predicted defrost time, the operation of the defrost element is maintained.

[0120] The start-up time of the defrost element can be determined by the determination method in the prior art, or can be predicted by a preset intelligent model.

[0121] like Figure 6 As shown, the processing device 20 further includes a training module 40 for training the prediction model.

[0122] like Figure 7 As shown, in one or more embodiments of the present application, the training module 40 includes a setting unit 401, an execution unit 402, and a generation unit 403. Each of these components can be implemented by a processor running a program.

[0123] In one or more embodiments of the present application, the setting unit 401 is configured to set the predicted network and the target network. The target network and the predicted network have the same network structure.

[0124] In one or more embodiments of the present application, the number of hidden layers in the prediction network, the number of neurons in each hidden layer, and the number of neurons in the input and output layers can be set according to specific application scenarios.

[0125] In one or more embodiments of the present application, the prediction network includes 3 hidden layers, each hidden layer has 64 neurons, and the input layer and output layer each have 16 neurons, which is adapted to the number of state parameters of the refrigerator. The architecture of the prediction network is relatively simple and is particularly suitable for realizing effective estimation of the state on a local processor.

[0126] The prediction network is denoted as Q e In one or more embodiments of the present application, the setting unit 401 is configured to randomly initialize the prediction network Q e Parameters

[0127] The target network is denoted as Q t In one or more embodiments of the present application, the setting unit 401 is configured to initialize the target network Q t Parameters satisfy

[0128] In one or more embodiments of the present application, the execution unit 402 is configured to obtain a first operating parameter of the refrigerator in the first working cycle. And input it into the prediction network, and adopt the ε-greedy strategy to select an optional defrost time as the defrost time to be evaluated.

[0129] More specifically, a probability value ε is set. In the initial state, ε is a relatively high value, such as 0.9. When selecting an optional defrost time as the defrost time to be evaluated, a random number is first generated, and the random number is between 0 and 1. If the random number is less than the probability value ε, an optional defrost time is randomly selected from the action space A; otherwise, the random number will be in the range of 1-ε, then based on the current knowledge and experience, the prediction network selects an optional defrost time that is considered to be optimal as the defrost time to be evaluated. As the number of training rounds decreases, the value of ε gradually decreases.

[0130] Based on current knowledge and experience, the prediction network is used to select the optimal optional defrost time. The prediction network will select the optional defrost time that can maximize the Q value under the current state, that is, a = argmax a∈|A| Q(s,a).

[0131] The execution unit 402 further drives the defrost element to operate, and turns off the defrost element when the operating time of the defrost element reaches the defrost time to be evaluated.

[0132] In one or more embodiments of the present application, the generating unit 403 is configured to obtain the second operating parameter of the refrigerator in the second working cycle. And input it into the prediction network, select the maximum Q value from the output of the prediction network, and generate a reward value.

[0133] In one or more embodiments of the present application, the reward value is generated based on the difference between the first operating parameter of the first working cycle refrigerator and the second operating parameter of the second working cycle refrigerator, and based on the energy consumption corresponding to the defrost time to be evaluated.

[0134] In one or more embodiments of the present application, the reward function can be expressed as:

[0135]

[0136] In the reward function, E a Represents the power consumption of the refrigerator when it is running according to the defrost time to be evaluated. Since energy consumption has a direct negative impact on the operating cost and efficiency of the refrigerator, the power consumption is set to a negative value in the reward function to encourage the prediction network to select an optional defrost time that can reduce energy consumption; It reflects the difference between the state of the refrigerator in the first working cycle and the second working cycle. If the difference between the two is large, the reward value is more negative, which encourages the prediction network to choose an optional defrosting time that keeps the state stable, ensuring that the state of the refrigerator will not fluctuate drastically during the defrosting process, avoiding affecting normal refrigeration.

[0137] like Figure 8As shown, in one or more embodiments of the present application, the training module 40 performs multiple steps as shown in the figure.

[0138] Step S201: Randomly initialize the prediction network Q e Parameters

[0139] Step S202: Initialize the target network Q t Parameters

[0140] Step S203: Obtaining the first operating parameter of the refrigerator in the first working cycle And input to the prediction network.

[0141] Step S204: adopting the ε-greedy strategy to select an optional defrost duration as the defrost duration to be evaluated.

[0142] Step S205: driving the defrosting element to operate.

[0143] Step S206: determining whether the operation time of the defrost element reaches the defrost time to be evaluated.

[0144] Step S207: If the operation time of the defrost element reaches the defrost time to be evaluated, the defrost element is turned off.

[0145] Step S208: If the operation time of the defrost element does not reach the defrost time to be evaluated, the operation of the defrost element is maintained.

[0146] Step S209: After the defrosting element is turned off, the second operating parameter of the refrigerator in the second working cycle is obtained. And input to the prediction network Q e .

[0147] Step S210: In the prediction network Q e Select the maximum Q value among the outputs.

[0148] Step S211: generating a reward value based on a difference between a first operating parameter of the refrigerator with a first working cycle and a second operating parameter of the refrigerator with a second working cycle, and based on energy consumption corresponding to the defrosting time to be evaluated.

[0149] In one or more embodiments of the present application, the training module 40 performs the following steps: Fig. 9 Multiple steps shown.

[0150] Step S301: Randomly initialize the prediction network Q e Parameters

[0151] Step S302: Initialize the target network Q t Parameters

[0152] Step S303: Determine whether the operating time of the refrigerator exceeds the preset frosting cycle time t1.

[0153] Step S304: If the working time of the refrigerator exceeds the preset frosting cycle time t1, the first operating parameter of the refrigerator in the first working cycle is obtained. And input to the prediction network Q e .

[0154] Step S305: If the working time of the refrigerator does not exceed the preset frosting cycle time t1, the refrigerator is kept in normal operation for the accumulated working time.

[0155] Step S306: adopting the ε-greedy strategy to select an optional defrost duration as the defrost duration to be evaluated.

[0156] Step S307: driving the defrost element to operate.

[0157] Step S308: determining whether the operation time of the defrost element reaches the defrost time to be evaluated.

[0158] Step S309: If the operation time of the defrost element reaches the defrost time to be evaluated, the defrost element is turned off.

[0159] Step S310: If the operation time of the defrost element does not reach the defrost time to be evaluated, the operation of the defrost element is maintained.

[0160] Step S311: After the defrosting element is turned off, the second operating parameter of the refrigerator in the second working cycle is obtained. And input to the prediction network Q e .

[0161] Step S312: In the prediction network Q e Select the maximum Q value among the outputs.

[0162] Step S313: generating a reward value based on a difference between a first operating parameter of the refrigerator with the first working cycle and a second operating parameter of the refrigerator with the second working cycle, and based on energy consumption corresponding to the defrosting time to be evaluated.

[0163] The duration of the frosting cycle t1 is related to the indoor ambient temperature, indoor ambient humidity, and the moisture content of the items in the storage compartment. The state data collected after the frosting cycle duration t1 is met is more closely related to the formation of the frost layer, which can make the prediction model better adapt to different usage scenarios, environmental changes, the relationship between usage habits and frost layer formation, and help to more accurately predict the optimal defrosting time.

[0164] In one or more embodiments of the present application, the training module 40 performs the following steps: Fig.10 Multiple steps shown.

[0165] Step S401: Randomly initialize the prediction network Q e Parameters

[0166] Step S402: Initialize the target network Q t Parameters

[0167] Step S403: Determine whether the operating time of the refrigerator exceeds the preset frosting cycle time t1.

[0168] Step S404: If the working time of the refrigerator exceeds the preset frosting cycle time t1, the first operating parameter of the refrigerator in the first working cycle is obtained. And input to the prediction network Q e .

[0169] Step S405: If the working time of the refrigerator does not exceed the preset frosting cycle time t1, the refrigerator is kept operating normally for the accumulated working time.

[0170] Step S406: adopting the ε-greedy strategy to select an optional defrost duration as the defrost duration to be evaluated.

[0171] Step S407: driving the defrost element to operate.

[0172] Step S408: determining whether the operation time of the defrost element reaches the defrost time to be evaluated.

[0173] Step S409: If the operation time of the defrost element reaches the defrost time to be evaluated, the defrost element is turned off.

[0174] Step S410: If the operation time of the defrost element does not reach the defrost time to be evaluated, the operation of the defrost element is maintained.

[0175] Step S411: After the defrost element is turned off, the operating time of the refrigerator is accumulated again.

[0176] Step S412: Determine whether the re-accumulated refrigerator operating time exceeds the preset evaluation cycle time t2.

[0177] The evaluation cycle duration t2 is the post-defrost cycle, and the operating parameters during this period can reflect the quality of the defrosting effect; if the defrosting time to be evaluated is not selected well, the defrosting will not be thorough, which will cause a series of problems, such as the temperature increase in the storage compartment, and these problems require a certain amount of time to be reflected, so the evaluation cycle duration t2 is specially set.

[0178] In one or more embodiments of the present application, the frosting cycle duration t1 is greater than the evaluation cycle duration t2.

[0179] Step S413: If the re-accumulated refrigerator operating time exceeds the preset evaluation cycle time t2, obtain the second operating parameter of the refrigerator in the second working cycle And input to the prediction network Qe.

[0180] Step S414: If the re-accumulated operating time of the refrigerator does not exceed the preset evaluation cycle time t2, the operation of the refrigerator is maintained until the re-accumulated operating time of the refrigerator exceeds the preset evaluation cycle time t2.

[0181] Step S415: In the prediction network Q e Select the maximum Q value among the outputs.

[0182] Step S416: generating a reward value based on a difference between a first operating parameter of the refrigerator with the first working cycle and a second operating parameter of the refrigerator with the second working cycle, and based on energy consumption corresponding to the defrosting time to be evaluated.

[0183] like Fig.11 As shown, in one or more embodiments of the present application, the training module 40 also includes a collection unit 404, a sampling unit 405, a target Q value calculation unit 406, a loss value calculation unit 407 and an update unit 408.

[0184] The collecting unit 404 collects sample data and stores the sample data in the experience pool D. Each piece of sample data includes: the first operating parameter of the refrigerator in the first working cycle, the defrosting time to be evaluated, the reward value, and the second operating parameter of the refrigerator in the second working cycle. The sample data is represented as Where a is the defrosting time to be evaluated, and r is the reward value.

[0185] The sampling unit 405 is configured to estimate whether the number of sample data in the experience pool D is higher than a preset number threshold N. If the number of sample data in the experience pool D is higher than the preset number threshold N, a batch of sample data corresponding to the preset number threshold N is extracted from the experience pool D.

[0186] The target Q value calculation unit 406 is configured as follows:

[0187] Set the first operating parameter of the refrigerator in the first working cycle in the batch sample data batch Input prediction network Q e , get the prediction network Q e The first maximum Q value of the output is denoted by q v .

[0188] The first maximum Q value q v satisfy:

[0189] Set the second operating parameters of the second working cycle refrigerator in the batch sample data batch Enter the target network Q t , get the target network Q t The second largest Q value of the output is denoted by q t .

[0190] The second largest Q value q t satisfy:

[0191] The target Q value is calculated based on the second maximum Q value, the reward value r in the batch sample data batch, and the discount factor γ.

[0192] Target Q value q t 'satisfy:

[0193] The loss value calculation unit 407 is configured to calculate the loss value according to the first maximum Q value q v and the target Q value q t 'Calculate the target network Q t With the prediction network Q e The loss value Loss.

[0194] The loss value Loss is expressed as:

[0195] Loss = MSE(q t ′,q v )

[0196] The updating unit 408 is configured to update the prediction network Q by back propagation e Configuration parameters.

[0197] The training module 40 performs the following steps: Fig.12 Multiple steps shown.

[0198] Step S501: Determine whether the number of sample data in the experience pool D is higher than a preset number threshold N.

[0199] Step S502: If the number of sample data in the experience pool D is higher than the preset number threshold N, a batch of sample data corresponding to the preset number threshold N is extracted from the experience pool D.

[0200] Step S503: If the amount of sample data in the experience pool D is not higher than the preset amount threshold N, the sample data in the experience pool D is accumulated.

[0201] Step S504: The first operating parameter of the refrigerator in the first working cycle in the batch sample data batch Input prediction network Qe , get the prediction network Q e The first maximum Q value q of the output v .

[0202] Step S505: The second operating parameter of the refrigerator in the second working cycle in the batch sample data batch Enter the target network Q t , get the target network Q t The second largest Q value q of the output t .

[0203] Step S506: Calculate the target Q value according to the second maximum Q value, the reward value r in the batch sample data batch, and the discount factor γ.

[0204] Step S507: According to the first maximum Q value q v and the target Q value q t 'Calculate the target network Q t With the prediction network Q e The loss value Loss.

[0205] Step S508: Update the prediction network Q through back propagation e Configuration parameters

[0206] Update the prediction network Q through back propagation e Configuration parameters An algorithm disclosed in the prior art is used.

[0207] like Fig.13 As shown, in one or more embodiments of the present application, the training module 40 further includes a counting unit 409 and a judging unit 410 .

[0208] The counting unit 409 is configured to update the prediction network Q in the updating unit 408 e After the configuration parameters are set, the count value step of the training counter is increased by 1.

[0209] The judgment unit 410 is configured to update the target network Q when the count value of the training counter can be divided by the preset freezing round M. t Configuration parameters.

[0210] like Fig.14 As shown, the training module 40 performs a plurality of steps as shown.

[0211] Step S601: Update the prediction network Q e Configuration parameters The count value of the training counter step is increased by 1.

[0212] Step S602: Determine whether the count value of the training counter is divisible by the preset freezing round M.

[0213] Step S603: If the count value step of the training counter can be divided by the preset freezing round M, the target network Q is updated. t Configuration parameters, namely

[0214] During the training process of the training module 40, the prediction network Q e The configuration parameters of are continuously updated to better approximate the optimal Q value function. e Correspondingly, due to the setting of training counter and preset freezing rounds, the target network Q t The configuration parameters are only updated from the prediction network Q at a certain number of steps. e Copied, updated less frequently, target network Q t The Q value function is relative to the prediction network Q e In this process, the target network Q t It is used to improve the stability of training, which is equivalent to a stabilizer. If the training module 40 only uses the prediction network Q e Select actions and calculate target Q values, which will change continuously, causing the training process to oscillate or even diverge. t This will make the target Q value relatively stable, which will help predict more stable learning of the network.

[0215] like Fig.15 As shown, in one or more embodiments of the present application, the training module 40 further includes an exit unit 411 .

[0216] The exit unit 411 is configured to exit the training when the loss value is lower than the preset training error or the count value of the training counter exceeds the preset total training round eps, and use the prediction network Q e as a predictive model.

[0217] The training module 40 performs the following steps: Fig.16 Multiple steps shown.

[0218] Step S701: Determine whether Loss≤E or step≥eps is satisfied;

[0219] Step S702: If Loss ≤ E or step ≥ eps is satisfied, the training is terminated to obtain the trained prediction network Q e Configuration parameters

[0220] In practical applications, it contains the latest learning results and can more accurately reflect the prediction network Q value in the current environment. eIt is deployed as a prediction model in the processing device 20 to perform prediction.

[0221] A flow chart of the training module 40 is shown in FIG. Fig.17 shown.

[0222] In the description of the above embodiments, specific features, structures, materials or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0223] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. Refrigerator, including: Evaporator; a fan for guiding the air heat-exchanged with the evaporator into the storage compartment; and a defrost element for melting frost on the evaporator; It is characterized by further comprising: A processing device comprising: A storage unit storing a prediction model constructed based on a deep Q network, wherein the state space of the prediction model includes operating parameters of the refrigerator, and the action space includes a plurality of optional defrosting durations; An acquisition unit, which is used to acquire operating parameters of the refrigerator in the current working cycle; a selection unit configured to input the operating parameters of the refrigerator in the current working cycle into the prediction model, and select the optional defrosting time with the largest Q value from the output of the prediction model as the predicted defrosting time; the Q value represents the expected value of executing one of the optional defrosting time under the operating parameters of the refrigerator in the current working cycle; and The control unit is configured to turn off the defrost element when the operation time of the defrost element reaches the predicted defrost time.

2. The refrigerator according to claim 1, characterized in that: The processing device also includes: A training module, which is used to train the prediction model; It includes: A setting unit configured to set a prediction network and a target network; an execution unit configured to obtain a first operating parameter of the refrigerator in the first working cycle and input the parameter into the prediction network, select an optional defrost duration as the defrost duration to be evaluated by adopting an ε-greedy strategy; drive the defrost element to operate, and close the defrost element when the operating duration of the defrost element reaches the defrost duration to be evaluated; and The generating unit is configured to obtain the second operating parameter of the refrigerator with the second working cycle and input it into the prediction network, select the maximum Q value from the output of the prediction network, and generate a reward value to obtain sample data.

3. The refrigerator according to claim 2, characterized in that: In the training state, the execution unit is configured to determine whether the working time of the refrigerator exceeds the preset frosting cycle time; when the working time of the refrigerator exceeds the frosting cycle time, obtain the first operating parameters of the refrigerator in the first working cycle and input them into the prediction network.

4. The refrigerator according to claim 3, characterized in that: In the training state, the generating unit is configured to re-accumulate the working time of the refrigerator after turning off the defrosting element, and determine whether the re-accumulated working time of the refrigerator exceeds the preset evaluation cycle time; When the re-accumulated refrigerator operating time exceeds the preset evaluation cycle time, obtaining a second operating parameter of the refrigerator in a second operating cycle and inputting it into the prediction network; The frosting cycle duration is longer than the evaluation cycle duration.

5. The refrigerator according to claim 4, characterized in that: The training module also includes: A collection unit configured to collect sample data and store the sample data in an experience pool D, each piece of sample data including: a first operating parameter of a refrigerator in a first working cycle, a defrosting time to be evaluated, a reward value, and a second operating parameter of a refrigerator in a second working cycle; A sampling unit configured to estimate whether the number of sample data in the experience pool is higher than a preset number threshold; if the number of sample data in the experience pool is higher than the preset number threshold, then extracting batch sample data corresponding to the preset number threshold from the experience pool; a target Q value calculation unit, configured to input a first operating parameter of a refrigerator with a first working cycle in the batch sample data into a prediction network to obtain a first maximum Q value output by the prediction network; input a second operating parameter of a refrigerator with a second working cycle in the batch sample data into a target network to obtain a second maximum Q value output by the target network, and calculate a target Q value according to the second maximum Q value, the reward value in the batch sample data, and a discount factor; a loss value calculation unit configured to calculate a loss value of a target network and a prediction network according to the first maximum Q value and the target Q value; and An updating unit is configured to update the configuration parameters of the prediction network through back propagation.

6. The refrigerator according to claim 5, characterized in that: The training module also includes: a counting unit configured to increase the count value of the training counter by 1 after the updating unit updates the configuration parameters of the prediction network; and The judging unit is configured to update the configuration parameters of the target network when the count value of the training counter can be divided by the preset freezing round.

7. The refrigerator according to claim 6, characterized in that: The training module also includes: An exit unit is configured to exit training when the loss value is lower than a preset training error or the count value of the training counter exceeds a preset total training round; and use the prediction network as the prediction model.

8. The refrigerator according to any one of claims 2 to 7, characterized in that: The reward value is generated based on a difference between a first operating parameter of the first working cycle refrigerator and a second operating parameter of the second working cycle refrigerator, and based on energy consumption corresponding to the defrosting time to be evaluated.

9. The refrigerator according to any one of claims 1 to 7, characterized in that: The operating parameters of the working cycle are the average values ​​of the operating parameters sampled within the working cycle.

10. The refrigerator according to any one of claims 1 to 7, characterized in that: The operating parameters of the working cycle include one or more of indoor ambient temperature, storage compartment set temperature, storage compartment actual temperature, evaporator surface temperature, compressor frequency and compressor power.

Images