A control method, electronic equipment, and air conditioning system for a residential water heater.

Abstract

This invention discloses a control method, electronic device, and air conditioning system for a residential water purifier. The control method includes: virtually adjusting the current indoor set temperature using a first control model to obtain a virtual value for the indoor set temperature; using this virtual value as a virtual control condition for a new indoor set temperature; using a second control model to predict the indoor ambient temperature and power consumption of the residential water purifier to obtain a predicted indoor ambient temperature t1 and a predicted power consumption P1; determining whether t1 meets the temperature setting condition and whether P1 meets the power consumption setting condition; and when t1 meets the temperature setting condition and P1 meets the power consumption setting condition, using the virtual value for the indoor set temperature as the new indoor set temperature to control the operation of the residential water purifier. This invention can better reduce power consumption when the temperature difference is small, achieving an energy-saving function suitable for every different home environment.

Description

Technical Field

[0001] This invention relates to the field of residential water purifier technology, and more particularly to a control method, electronic equipment, and air conditioning system for a residential water purifier. Background Technology

[0002] With the development of science and technology and a deeper understanding of energy conservation, along with the integration of smart home visualization technology into people's daily lives, people have placed higher demands on energy conservation and temperature adaptation. Currently, users of residential water chillers primarily set the temperature manually, and these settings are often not changed for extended periods. This method is essentially a form of constant temperature control, only ensuring that the indoor temperature approaches the set temperature, and cannot adjust the set temperature according to environmental changes. Therefore, some have proposed a power consumption optimization scheme for residential water chillers based on reinforcement learning. Similar schemes exist for temperature-controlled appliances such as air conditioners and heaters. For example, a related patent discloses a method that uses a pre-trained reinforcement learning model to obtain the control parameters at the current moment, and uses these control parameters to control the operating state of temperature regulation equipment in a data center until the temperature in the data center stabilizes. However, because it employs deep learning (reinforcement learning), the inference results often have to be performed in the cloud, requiring significant computational costs. In some cases, it is even necessary to model different users individually to achieve the best results, resulting in even greater computational costs. Summary of the Invention

[0003] In view of this, the present invention discloses a control method, electronic equipment and air conditioning system for a household water heater, in order to solve the problem of poor energy-saving effect of existing household water heaters.

[0004] To achieve the above objectives, the technical solution adopted by this invention is as follows:

[0005] The first aspect of this invention discloses a control method for a residential water purifier, wherein the residential water purifier is pre-configured with a first control model and a second control model, and the method includes:

[0006] The current indoor set temperature is virtually adjusted using the first control model to obtain a virtual value for the indoor set temperature.

[0007] The virtual value of the indoor set temperature is used as the virtual control condition for the new indoor set temperature. The second control model is used to predict the indoor ambient temperature and power consumption of the household water heater, and the predicted indoor ambient temperature t1 and predicted power consumption P1 of the household water heater are obtained.

[0008] Determine whether t1 meets the temperature setting condition and whether P1 meets the power consumption setting condition;

[0009] When t1 meets the temperature setting condition and P1 meets the power consumption setting condition, the virtual value of the indoor set temperature is used as the new indoor set temperature to control the operation of the household water heater.

[0010] Further, optionally, the current indoor set temperature is virtually adjusted using the first control model to obtain a virtual value for the indoor set temperature, including:

[0011] Acquire real-time status data and corresponding environmental data for the residential water purifier, including the current indoor set temperature;

[0012] The real-time status data and corresponding environmental data are input into the first control model, and multiple adjustment action estimates are output. Each adjustment action estimate corresponds to a temperature adjustment action, which includes heating, cooling or keeping the temperature constant.

[0013] Select the adjustment action valuation that meets the set requirements from multiple adjustment action valuations;

[0014] The virtual value of the indoor set temperature is determined based on the estimated adjustment actions that meet the set requirements.

[0015] Optionally, select the adjustment action estimate that meets the set requirements from multiple adjustment action estimates, and adjust the indoor set temperature, including:

[0016] Select the maximum value from multiple adjustment action estimates, and adjust the indoor set temperature according to the temperature adjustment action corresponding to the maximum value.

[0017] Further, optionally, the current indoor set temperature is virtually adjusted using the first control model to obtain a virtual value for the indoor set temperature, including:

[0018] Input real-time status data and corresponding environmental data into the first control model, and output the virtual value of the indoor set temperature.

[0019] Further optionally, before using the virtual value of the indoor set temperature as the virtual control condition for the new indoor set temperature, and using the second control model to predict the indoor ambient temperature and power consumption of the household water heater to obtain the predicted indoor ambient temperature t1 and the predicted power consumption P1 of the household water heater, the method further includes:

[0020] The second control model is trained using historical status data of the user-type water turbine and corresponding environmental data.

[0021] Optionally, the second control model is trained using historical state data of the user-type water turbine and corresponding environmental data to obtain the second control model, which includes:

[0022] Input the status data of the household water purifier and the corresponding environmental data collected at each of the current time and the past M times into the second control model, and output the predicted value of the indoor ambient temperature t0 at the next time, where M≥2;

[0023] Obtain the indoor ambient temperature t0' at the next moment;

[0024] Update the model parameters of the second control model based on the comparison results of t0 and t0';

[0025] Repeat the training until the training is completed.

[0026] Optionally, the virtual value of the indoor set temperature is used as the virtual control condition for the new indoor set temperature. The second control model is then used to predict the indoor ambient temperature and power consumption of the residential water heater, resulting in the predicted indoor ambient temperature t1 and the predicted power consumption P1 of the residential water heater, including:

[0027] Acquire the status data of the household water purifier and the corresponding environmental data collected at each of the current time and the M previous times; replace the current indoor set temperature with the virtual value of the indoor set temperature and input it into the trained second control model, and output the predicted indoor environmental temperature t1 and the predicted power consumption P1 of the household water purifier at the next time.

[0028] Further, optionally, determining whether the predicted indoor ambient temperature t1 meets the temperature setting conditions, and determining whether the predicted power consumption P1 of the residential water chiller meets the power consumption setting conditions, includes:

[0029] The second control model is used to predict the indoor ambient temperature and power consumption of the household water heater under the current indoor set temperature control conditions, and the predicted indoor ambient temperature t2 and predicted power consumption P2 of the household water heater are obtained.

[0030] Compare t1 and t2, and compare P1 and P2;

[0031] When |t1-t2|≤a, a is a natural number greater than 0, which is considered to satisfy the temperature setting condition;

[0032] When P1≤P2, the power consumption setting condition is considered met.

[0033] Further optionally, if t1 does not meet the temperature setting condition, and / or P1 does not meet the power consumption setting condition, the method further includes:

[0034] The model parameters of the first control model and the model parameters of the second control model are updated based on the comparison results.

[0035] Further optionally, the first control model is a DQN model and the second control model is an LSTM model; the first and second control models are deployed in a local embedded module.

[0036] Further optionally, the status data of the residential water chiller includes at least one of the following: inlet water temperature, outlet water temperature, compressor power consumption, suction temperature, exhaust temperature, indoor fan power consumption, and outdoor fan power consumption.

[0037] A second aspect of the present invention discloses an electronic device, the electronic device comprising:

[0038] Memory, used to store computer instructions;

[0039] A controller is used to invoke and execute computer instructions stored in memory to implement the methods provided in any of the first aspects.

[0040] A third aspect of the present invention discloses an air conditioning system that employs the method provided in any of the first aspects; or, the air conditioning system includes electronic equipment as provided in the second aspect.

[0041] Beneficial effects: The household water purifier of the present invention is equipped with a first control model and a second control model. Based on the first control model, the current indoor set temperature can be virtually adjusted. Based on the second control model, the indoor ambient temperature t1 and power consumption P1 of the household water purifier under the virtual value of the indoor set temperature can be predicted and used as a constraint to update the indoor set temperature, so as to better reduce power consumption when the temperature difference is not large, and achieve the function of being suitable for each different home environment and saving energy. Attached Figure Description

[0042] The above and other objects, features, and advantages of the present invention will become more apparent from the detailed description of exemplary embodiments with reference to the accompanying drawings. The drawings described below are merely some embodiments disclosed in the present invention; those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0043] Figure 1 A schematic diagram of a control method for a household water dispenser according to an embodiment of the present invention is shown as an example.

[0044] Figure 2 A schematic diagram of a control method for a household water dispenser according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. “Multiple” generally includes at least two, but does not exclude the inclusion of at least one.

[0047] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0048] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes said element.

[0049] To further illustrate the technical solutions of this invention, the "xx model" appearing in the relevant embodiments of this invention will first be explained. The first control model and the second control model proposed in this invention can be understood as the mapping relationship / functional relationship between input parameters and output parameters, such as long short-term memory models, neural network models, formulas, mapping tables, etc.

[0050] The first aspect of this invention discloses a control method for a household water purifier, wherein the household water purifier is pre-configured with a first control model and a second control model, such as... Figure 1 As shown, the method includes S1 to S4, wherein:

[0051] S1, Use the first control model to virtually adjust the current indoor set temperature to obtain a virtual value of the indoor set temperature;

[0052] S2, use the virtual value of the indoor set temperature as the virtual control condition for the new indoor set temperature, and use the second control model to predict the indoor ambient temperature and power consumption of the household water heater, so as to obtain the predicted indoor ambient temperature t1 and the predicted power consumption of the household water heater P1.

[0053] S3, determine whether t1 meets the temperature setting condition and whether P1 meets the power consumption setting condition; when t1 meets the temperature setting condition and P1 meets the power consumption setting condition, execute S4.

[0054] S4 uses the virtual value of the indoor set temperature as the new indoor set temperature to control the operation of the household water heater.

[0055] The first control model in this embodiment can adjust the indoor set temperature, but this adjustment is not a real adjustment; it is a virtual adjustment. The second control model predicts the indoor ambient temperature and power consumption of the water purifier under the control condition of the virtual indoor set temperature. It further judges whether the predicted indoor ambient temperature and predicted power consumption meet the set conditions of constant indoor temperature and reduced power consumption. If they meet, the current indoor set temperature is updated, and the virtual indoor set temperature is used as the new indoor set temperature to control the operation of the water purifier. If not, the model parameters of the first and second control models are adjusted. This allows these two models to adapt to each household and each user. Compared with traditional methods, this not only ensures constant temperature control but also reduces energy consumption, thus solving the problem of poor energy efficiency in existing water purifiers and their inability to adjust the set temperature according to environmental changes.

[0056] The update process of the two models mentioned above is existing technology, and this embodiment does not impose specific limitations on it.

[0057] Preferably, the two models described above are deployed on a local embedded module. Data from the user is collected on the embedded device, and the model corrects the deviation between the model and the expected value (predicted value) through self-learning. Then, actions are generated based on reinforcement learning to reduce power consumption, making it suitable for various home environments. Simultaneously, self-learning and model optimization on the embedded device only requires collecting a few days of indoor temperature data from the user locally to optimize the power consumption of the residential water purifier. This eliminates reliance on cloud computing and removes the communication module, resulting in significant cost savings compared to traditional methods.

[0058] Furthermore, in S1, the first control model is used to virtually adjust the current indoor set temperature to obtain a virtual value for the indoor set temperature, including S11 to S14, wherein:

[0059] S11, obtain real-time status data of the household water heater and corresponding environmental data, including the current indoor set temperature;

[0060] S12, input the real-time status data and the corresponding environmental data into the first control model, and output multiple adjustment action estimates. Each adjustment action estimate corresponds to a temperature adjustment action, which includes heating, cooling or keeping the temperature constant.

[0061] S13, Select the adjustment action valuation that meets the set requirements from multiple adjustment action valuations;

[0062] S14, determine the virtual value of the indoor set temperature based on the estimated adjustment action that meets the set requirements.

[0063] In this embodiment, the first control model processes the real-time status data of the residential water heater and the corresponding environmental data. For example, at 4:50, a set of residential water heater status data and corresponding indoor and outdoor environmental data are collected using sensors, including indoor and outdoor temperatures, inlet water temperature, outlet water temperature, indoor set temperature, compressor power consumption, compressor air pressure, fan power consumption, and overall power consumption of the residential water heater. This set of data serves as the input parameters for the first control model, and its output parameters are multiple adjustment action estimates. Each adjustment action estimate corresponds to an action that adjusts the indoor set temperature. For heating, each adjustment action estimate preferably corresponds to a different cooling action or maintaining a constant temperature; for cooling, each adjustment action estimate preferably corresponds to a different heating action or maintaining a constant temperature. An adjustment action estimate that meets the set requirements is selected from the multiple adjustment action estimates to adjust the indoor set temperature, thereby obtaining a virtual value for the indoor set temperature. For example, the estimate for a 1°C temperature increase is 100, and the estimate for a 3°C temperature increase is 200.

[0064] The first control model (reinforcement learning model) in this embodiment uses the current state and environment of the household water heater as input data. During model training, it ensures that the indoor temperature change caused by its operation differs from the temperature without reinforcement learning by less than 1 degree Celsius, while minimizing power consumption. Simultaneously, it aims to bring the temperature closer to the predicted temperature of the second control model, thereby ensuring the desired effect is achieved.

[0065] Optionally, S13 selects the adjustment action valuation that meets the set requirements from multiple adjustment action valuations, including S131 to S132, wherein:

[0066] S131, Select the maximum value from multiple adjustment estimates;

[0067] S132, adjust the current indoor set temperature according to the adjustment action corresponding to the maximum value.

[0068] With energy saving as the goal, the maximum value among multiple adjustment values ​​is preferred, which corresponds to a larger temperature adjustment range. Taking heating as an example, the value of a 1°C temperature drop is 100, the value of a 2°C temperature drop is 200, and the value of a 3°C temperature drop is 300. Therefore, a 3°C temperature drop is preferred, which can achieve both temperature assurance and greater energy saving.

[0069] This embodiment also provides another implementation method for calculating the virtual value of the indoor set temperature, specifically:

[0070] The real-time collected status data of the household water heater and the corresponding environmental data are input into the first control model, and the virtual value of the indoor set temperature is output.

[0071] This embodiment uses the virtual value of the indoor set temperature as the output parameter to optimize the running program of the first control model, which helps to reduce the power consumption of the whole machine.

[0072] Furthermore, before using the virtual value of the indoor set temperature as the virtual control condition for the new indoor set temperature in S2, and using the second control model to predict the indoor ambient temperature and power consumption of the household water heater, and obtaining the predicted indoor ambient temperature t1 and predicted power consumption P1 of the household water heater, the method further includes step A1:

[0073] A1. The second control model is trained using the historical status data of the user-type water turbine and the corresponding environmental data to obtain the second control model.

[0074] In this preferred embodiment, the second control model is pre-trained to obtain a trained first control model. The trained second control model is used to predict the power consumption and indoor ambient temperature of the household water heater at the current moment and the next moment at the original indoor set temperature, as well as the power consumption and indoor ambient temperature at the new indoor set temperature. This improves prediction accuracy, better meets user temperature requirements, and saves energy.

[0075] Furthermore, A1 includes A11 to A14, where:

[0076] A11 inputs the status data of the household water purifier collected at the current time and at each of the past M times, along with the corresponding environmental data, into the second control model, and outputs the predicted value t0 of the indoor ambient temperature at the next time.

[0077] A12, obtain the indoor ambient temperature t0' at the next moment;

[0078] A13, Update the model parameters of the second control model based on the comparison results of t0 and t0';

[0079] A14, repeat training until the training ends.

[0080] For example, with a data collection time interval of 1 hour, M = 3, and the current time being 5 o'clock, a set of status data for a household water heater and corresponding indoor and outdoor environmental data are collected at 2 o'clock, 3 o'clock, 4 o'clock, and 5 o'clock respectively. This data is input into an LSTM model, outputting a predicted indoor environmental temperature t0 at 6 o'clock. The indoor environmental temperature t0' at 6 o'clock is then obtained, and the deviation between the actual value t0' and the predicted value t0 is used to correct the second control model. Specifically, if |t0 – t0'| ≤ b, where b is a natural number and 0 ≤ b ≤ 1, then training is complete; if |t0 – t0'| > b, then the model parameters of the second control model are adjusted based on the comparison between t0 and t0'. After adjustment, the next round of training is performed, using the data collected at 3 o'clock, 4 o'clock, 5 o'clock, and 6 o'clock as input parameters into the second control model to predict the indoor environmental temperature at 7 o'clock. This training process is repeated until the training is complete.

[0081] Furthermore, S2 includes:

[0082] The system acquires the status data of the household water purifier and the corresponding environmental data collected at each of the current time and the M previous times. It replaces the current indoor set temperature with the virtual value of the indoor set temperature and inputs it into the second control model. The system outputs the predicted indoor environmental temperature t1 and the predicted power consumption P1 of the household water purifier for the next time.

[0083] For example, the input to an LSTM model is 4 hours of data on the status and environment of a residential water purifier, with dimensions [4,10], referring to the data for each hour within those 4 hours, including compressor pressure, set temperature, etc. The output is a prediction of the indoor temperature and power consumption for the 5th hour. The input to a DQN model is real-time environmental status data [1,10], and the output is the adjustment action (heating up, cooling down, or keeping it unchanged).

[0084] Specifically, at 5 o'clock, the LSTM predicts based on the data from the previous 4 hours that the set temperature temp1 will continue to be maintained. Therefore, the indoor temperature at 5 o'clock will be t2, and the power consumption will be p2 (this t2 is compared with the actual environment at 5 o'clock and is used to update the LSTM model to ensure the accuracy of the model).

[0085] At this point, the DQN model generates an action (adjustment action estimate) for the environment at point 4:50, thereby virtually adjusting the current indoor ambient temperature to obtain a virtual indoor set temperature value. This virtual indoor set temperature value replaces the indoor set temperature at point 5 (i.e., the current indoor set temperature), and is used again as the input to the LSTM (the input at this time is the data from points 2, 3, 4, and 4:50), yielding t1 and p1. The difference between t1 and t2 is used to determine whether the effect of keeping the temperature constant has been achieved and whether power consumption has been reduced (the difference between p1 and p2), thus determining whether the set temperature generated by this DQN should be used at point 5.

[0086] Furthermore, S3 includes S31 to S34, wherein:

[0087] S31, using the second control model, the indoor ambient temperature and power consumption of the household water heater under the current indoor set temperature control conditions are predicted to obtain the predicted indoor ambient temperature t2 and the predicted power consumption of the household water heater P2.

[0088] S32, compare t1 and t2, and compare P1 and P2;

[0089] S33, when |t1-t2|≤a, a is a natural number greater than 0, which is considered to satisfy the temperature setting condition;

[0090] S34, when P1≤P2, is considered to meet the power consumption setting condition.

[0091] This embodiment further restricts the constraints mentioned above. Under the premise of constant temperature and energy saving, |t1-t2≤a is satisfied between t1 and t2. The value of b is between 0 and 2℃, such as 0.5, 1, 1.5, 2, etc. Only when both temperature and power consumption meet the conditions will the indoor set temperature for the next moment be adjusted. That is, the operation of the household water purifier is controlled by the new indoor set temperature.

[0092] Preferably, the first control model is a DQN model and the second control model is an LSTM model; the first and second control models are deployed in a local embedded module.

[0093] The first control model obtains the virtual value of the indoor set temperature based on the real-time status data of the household water purifier and the corresponding environmental data, that is, it obtains the virtual value of the indoor set temperature based on the status data of the household water purifier at the current moment and the corresponding environmental data. The second control model obtains the predicted indoor environmental temperature and the predicted power consumption of the household water purifier based on the historical status data of the household water purifier and the corresponding environmental data, that is, it obtains the predicted indoor environmental temperature and the power consumption of the household water purifier at the next moment based on the status data of the household water purifier at the current moment and the M moments before it and the corresponding environmental data.

[0094] This embodiment provides an embedded-based method that deploys both the LSTM and DQN models in an embedded module, eliminating the reliance on cloud computing, removing the communication module, and reducing the cost of later algorithm maintenance by 50%.

[0095] Optionally, the status data of the residential water chiller includes at least one of the following: inlet water temperature, outlet water temperature, compressor power consumption, suction temperature, exhaust temperature, indoor fan power consumption, and outdoor fan power consumption. Environmental data also includes indoor ambient temperature and outdoor ambient temperature. The status data of the residential water chiller includes real-time status data and historical status data.

[0096] In one specific embodiment, the control method for a household water dispenser includes the following steps:

[0097] Step 1: Use LSTM to predict the indoor ambient temperature and the power consumption of the household water purifier at the next time step, and obtain the predicted indoor ambient temperature t1 and the predicted power consumption of the household water purifier P1.

[0098] Step 2: Use DQN to virtually adjust the current indoor set temperature to obtain a virtual value for the indoor set temperature;

[0099] Step 3: Input the virtual value of the indoor set temperature into the LSTM to replace the current indoor set temperature in the LSTM input parameters, and output the predicted indoor environmental temperature t2 and the predicted power consumption P2 of the household water purifier at the next time step.

[0100] Step 4: Compare t1 and t2, and compare P1 and P2;

[0101] Step 5: When |t1-t2|≤a, a is a natural number greater than 0, and P1≤P2, the virtual value of the indoor set temperature is used as the new indoor set temperature to control the operation of the household water heater.

[0102] Step 6: When |t1-t2|>a, and / or P1>P2, update the model parameters of LSTM and DQN based on the comparison results.

[0103] The following is combined with Figure 2 The method of this embodiment will be described in further detail.

[0104] 1. The first step is to establish a unified large model. Using the daily data of 1,000 users (data including hourly set temperature, inlet water temperature, indoor temperature, outdoor temperature, compressor power consumption, air pressure and overall power consumption) as training data, two basic models are established: LSTM (Long Short-Term Memory) and DQN (Deep Q-Learning).

[0105] 2. The LSTM model is trained using data from previous user settings at different times, outdoor temperatures, and set temperatures. The goal is to accurately predict the user's set temperature and corresponding indoor temperature for the next few hours, thus providing guidance for subsequent reinforcement learning.

[0106] 3. The DQN model uses the current state and environment of the residential water purifier as input data. During model training, it ensures that the indoor temperature change caused by the purifier's operation differs from the temperature without reinforcement learning by less than 1 degree Celsius, while minimizing power consumption. Simultaneously, it aims to approximate the temperature predicted by the LSTM, thereby achieving the user's desired effect. The algorithm is shown below:

[0107] 3.1 Using the current environment as the input to the LSTM, predict the future set temperature;

[0108] 3.2. Starting from time t0, the environment at this time is denoted as s0. s0 is input into DQN to obtain the value assessment of action a0 (action: heating up, cooling down, unchanged) in this environment. Based on action a0, the corresponding reward r0 is further obtained. The action value function is... It can be expressed as (where the discount rate γ∈[0,1]);

[0109]

[0110] 3.3. Inputting s0 into the DQN yields an estimated value q0. Then, a0 generates a new environment s1 and a corresponding return r0. Inputting s1 into the DQN again yields a new estimated value q1. Therefore, we can consider s1 to be more accurate than s0, thus introducing an error:

[0111]

[0112] 3.4 Calculate the gradient of the loss function and update the model parameters w;

[0113]

[0114] In practice, implementing this on a board (embedded) platform requires considering many factors. The model constructed earlier was based on data from 1000 users, but if ported to an embedded system, its application to a specific user would have significant discrepancies (due to differences in household size, building materials, etc.). Therefore, the online learning method provided in this embodiment is necessary.

[0115] 4. Terminal implementation steps: Collect user data from the past few days, compare the data after model inference with the actual environmental information (the data comparison mainly involves two aspects: comparison between actual temperature and predicted temperature; comparison between power consumption at the previous moment and power consumption after changing the set temperature), in order to complete the gradient calculation of the embedded model, thereby updating the specific parameters of the model to adapt to different home environments.

[0116] Ultimately, the desired temperature predicted by LSTM is used as an indicator to change the indoor set temperature, thereby changing the indoor temperature and achieving the goal of both ensuring temperature and saving energy.

[0117] The second aspect of this embodiment discloses an electronic device, which includes: a memory for storing computer instructions; and a controller for calling and executing the computer instructions stored in the memory to implement the method provided in any of the first aspects.

[0118] The third aspect of this embodiment discloses an air conditioning system that employs the method provided in any of the first aspects; or, the air conditioning system includes electronic equipment as provided in the second aspect.

[0119] This invention provides an energy-saving method for residential water purifiers based on embedded online learning. It collects partial user data on an embedded device, corrects the deviation between the model and the expected (predicted) values ​​through self-learning, and then generates actions based on reinforcement learning to reduce power consumption. This method is adaptable to each household environment. In practical applications, no additional computational overhead is required; data inference is completed solely through the embedded system, and power consumption is effectively reduced while maintaining a relatively constant temperature. It eliminates reliance on cloud computing, greatly protecting user data security, and removes the communication module, reducing the cost of subsequent algorithm maintenance by at least 50%.

[0120] In the different embodiments provided by this invention, the same parameters, terms, logic, etc. should be understood to have the same meaning, and this application does not intentionally repeat the description in each embodiment.

[0121] Exemplary embodiments of the present disclosure have been specifically shown and described above. It should be understood that the present disclosure is not limited to the detailed structures, arrangements, or implementation methods described herein; rather, the present disclosure is intended to cover various modifications and equivalent arrangements contained within the spirit and scope of the appended claims.

Claims

1. A control method of a domestic water machine, characterized by, The household water purifier is pre-configured with a first control model and a second control model, which are embedded modules deployed locally. The method includes: The current indoor set temperature is virtually adjusted using the first control model to obtain a virtual value for the indoor set temperature. The virtual value of the indoor set temperature is used as the virtual control condition for the new indoor set temperature. The second control model is used to predict the indoor ambient temperature and power consumption of the household water heater, so as to obtain the predicted indoor ambient temperature t1 and the predicted power consumption P1 of the household water heater. The second control model is used to predict the indoor ambient temperature and power consumption of the household water heater under the current indoor set temperature control conditions, and the predicted indoor ambient temperature t2 and predicted power consumption P2 of the household water heater are obtained. Determine whether t1 meets the temperature setting condition and whether P1 meets the power consumption setting condition. The temperature setting condition includes |t1-t2|≤a, where a is a natural number greater than 0. The power consumption setting condition includes P1≤P2. When t1 meets the temperature setting condition and P1 meets the power consumption setting condition, the virtual value of the indoor set temperature is used as the new indoor set temperature to control the operation of the household water purifier. When |t1-t2|>a, and / or P1>P2, update the model parameters of the first control model and the model parameters of the second control model based on the comparison results.

2. The method of claim 1, wherein, The step of virtually adjusting the current indoor set temperature using the first control model to obtain a virtual value for the indoor set temperature includes: Acquire the real-time status data and corresponding environmental data of the household water purifier, wherein the environmental data includes the current indoor set temperature; The real-time status data and the corresponding environmental data are input into the first control model, and multiple adjustment action estimates are output. Each adjustment action estimate corresponds to a temperature adjustment action, which includes heating, cooling or keeping the temperature constant. Select the adjustment action valuation that meets the set requirements from the multiple adjustment action valuations; The virtual value of the indoor set temperature is determined based on the estimated adjustment action that meets the set requirements.

3. The method of claim 2, wherein, The step of selecting an adjustment action estimate that meets the set requirements from the plurality of adjustment action estimates and adjusting the indoor set temperature includes: The maximum value is selected from the multiple adjustment action estimates, and the current indoor set temperature is adjusted according to the temperature adjustment action corresponding to the maximum value.

4. The method as described in claim 1, characterized in that, The step of virtually adjusting the current indoor set temperature using the first control model to obtain a virtual value for the indoor set temperature includes: The real-time status data of the household water purifier and the corresponding environmental data are input into the first control model, and the virtual value of the indoor set temperature is output.

5. The method as described in claim 1, characterized in that, Before using the second control model to predict the indoor ambient temperature and power consumption of the household water heater, and obtaining the predicted indoor ambient temperature t1 and predicted power consumption P1 of the household water heater, the method further includes: The second control model is trained using the historical status data of the household water heater and the corresponding environmental data to obtain the second control model.

6. The method as described in claim 5, characterized in that, The process of training the second control model using historical state data and corresponding environmental data of the household water dispenser to obtain the second control model includes: The status data of the household water purifier and the corresponding environmental data collected at each of the current time and the past M times are input into the second control model, and the predicted value of the indoor ambient temperature t0 at the next time is output, where M≥2; Obtain the indoor ambient temperature t0' at the next moment; The model parameters of the second control model are updated based on the comparison results of t0 and t0'. Repeat the training until the training is completed.

7. The method as described in claim 1, characterized in that, The step of using the virtual value of the indoor set temperature as a virtual control condition for the new indoor set temperature, and using the second control model to predict the indoor ambient temperature and power consumption of the household water heater to obtain the predicted indoor ambient temperature t1 and the predicted power consumption P1 of the household water heater includes: Acquire the status data and corresponding environmental data of the household water purifier collected at each of the current time and the M previous times; The current indoor set temperature is replaced with the virtual value of the indoor set temperature and input into the second control model. The predicted indoor environmental temperature t1 and the predicted power consumption P1 of the household water purifier are obtained at the current time, where M≥2.

8. The method as described in claim 1, characterized in that, The first control model is a DQN model, and the second control model is an LSTM model; the first control model and the second control model are deployed in a local embedded module.

9. The method as described in claim 1, characterized in that, The status data of the household water heater includes at least one of the following: inlet water temperature, outlet water temperature, compressor power consumption, suction temperature, exhaust temperature, indoor fan power consumption, and outdoor fan power consumption.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store computer instructions; A controller for invoking and executing computer instructions stored in the memory to implement the method as described in any one of claims 1-9.

11. An air conditioning system, characterized in that, The air conditioning system employs the method described in any one of claims 1-9; or, The air conditioning system includes the electronic device as described in claim 10.

Images