Refrigerator parameter real-time cloud regulation and control system based on streaming Bayesian optimization

Abstract

The invention provides a refrigerator parameter real-time cloud regulation and control system based on streaming Bayesian optimization, and the system comprises a data obtaining module which is configured to obtain target multi-dimensional parameters of a to-be-regulated and controlled refrigerator in real time; the flow processing module is configured to pre-process the target multi-dimensional parameters, construct a message queue according to the pre-processed multi-dimensional parameters and push the message queue; the multi-dimensional parameters in the message queue are parameters of the refrigerator to be regulated and controlled in an abnormal state; the parameter adjustment module is configured to perform optimal parameter search on the multi-dimensional parameters in the message queue based on Bayesian optimization to obtain target parameters; and the control execution module is configured to adjust operation parameters of hardware equipment in the refrigerator to be regulated and controlled according to the target parameters. Through the system, the problems that an existing refrigerator parameter adjusting technology is low in efficiency, and real-time dynamic adjustment cannot be achieved are solved.

Description

Technical Field

[0001] This invention relates to the field of refrigerator parameter adjustment technology, and in particular to a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization. Background Technology

[0002] As a common household appliance, the performance of a refrigerator's control system directly impacts food preservation, user experience, and household energy consumption. With the deepening of energy conservation, environmental protection, and smart home concepts, the market is placing higher demands on refrigerators in terms of temperature control accuracy, operational energy efficiency, and adaptive capabilities. However, traditional refrigerator control systems, and those currently with limited intelligence, still suffer from several inherent flaws in their architecture design and algorithm strategies, hindering further improvements in their overall performance. These flaws are mainly manifested in the following aspects:

[0003] Regarding the adaptability of control parameters, most existing systems adopt static or semi-static setting strategies. For example, key parameters such as compressor start-up and shutdown thresholds, defrosting cycles, and runtime are usually preset to fixed values ​​during the production phase, or only a few predefined modes (such as seasonal modes) are provided for users to manually switch between. This approach makes it difficult to make real-time, autonomous, and dynamic adjustments based on constantly changing ambient temperature and humidity, frequency of door opening, dynamic fluctuations in the heat load of stored items, and personalized user habits. This results in a continuous deviation between the control process and actual needs, affecting the stability of storage quality.

[0004] In terms of energy efficiency optimization, most existing control schemes rely on simple rules (such as timed start-stop) or classical proportional-integral-derivative control methods. While these methods can maintain a general internal temperature, they often struggle to achieve optimal overall system energy consumption while meeting stringent requirements for temperature uniformity and fluctuation range. The lack of refined and proactive control over the operating status of components such as compressors and fans easily leads to problems like frequent compressor start-stops, ineffective idling, or excessive operation, resulting in unnecessary energy loss and contradicting the trend of green energy conservation.

[0005] Regarding real-time control response, many algorithms designed to optimize operation employ offline or batch processing computation modes. These algorithms typically rely on historical data for analysis and decision-making, making it impossible to respond quickly to rapidly changing thermodynamic states and real-time signals from user operations during refrigerator operation. The significant delay between decision-making and execution means that control commands often lag behind actual demand changes, leading to problems such as temperature overshoot and prolonged adjustment cycles, thus greatly reducing control effectiveness.

[0006] In terms of system architecture and scalability, most existing technical solutions are limited to computation and decision-making by the device's local microprocessor. Constrained by the limited computing power, storage resources, and algorithm complexity of local hardware, it is difficult to deploy and run advanced intelligent control algorithms that require large-scale data training and complex model iteration. This isolation prevents the system from effectively utilizing the powerful computing capabilities and massive data resources of cloud or edge collaborative nodes for continuous training of complex models, in-depth analysis of user behavior, and global energy efficiency optimization through multi-device collaboration, creating a bottleneck in the intelligent upgrade path. Summary of the Invention

[0007] This application provides a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization to solve the problems of low efficiency and inability to dynamically adjust refrigerator parameters in real time.

[0008] The system includes:

[0009] A data acquisition module is installed on the refrigerator to be controlled, and the data acquisition module is configured to acquire the target multi-dimensional parameters of the refrigerator to be controlled in real time.

[0010] The process processing module is located in the cloud and is configured to preprocess the target multi-dimensional parameters, construct a message queue based on the preprocessed multi-dimensional parameters, and push the message queue; the multi-dimensional parameters in the message queue are the parameters of the refrigerator to be controlled in an abnormal state.

[0011] A parameter adjustment module, which is located in the cloud, is configured to perform optimal parameter search on the multi-dimensional parameters in the message queue based on Bayesian optimization to obtain the target parameters.

[0012] A control execution module is configured to adjust the operating parameters of the hardware devices in the refrigerator to be controlled according to the target parameters.

[0013] Preferably, the data acquisition module includes:

[0014] A temperature monitoring unit, configured to acquire refrigerator equipment temperature parameters at several locations within the refrigerator to be regulated;

[0015] A power monitoring unit, configured to acquire compressor power parameters of the compressor in the refrigerator to be regulated;

[0016] An environmental monitoring unit is configured to acquire the ambient temperature and humidity parameters of the environment in which the refrigerator to be regulated is located, as well as the refrigerator door status parameters.

[0017] Preferably, the data acquisition module further includes:

[0018] A data encryption unit is configured to encrypt the refrigerator device temperature parameters, the compressor power parameters, the ambient temperature parameters, the ambient humidity parameters, and the refrigerator door status parameters according to a first key to obtain the target multi-dimensional parameters.

[0019] A data sending unit is configured to report the target multi-dimensional parameters via the MQTT protocol.

[0020] Preferably, the process processing module includes:

[0021] A data decryption unit, configured to decrypt the target multi-dimensional parameters based on a first public key;

[0022] A data preprocessing unit is configured to preprocess the decrypted target multi-dimensional parameters to obtain preprocessed parameters; the preprocessing includes outlier filtering and data standardization.

[0023] The task flow construction unit is built on a distributed stream processing platform based on Apache Flink and is configured to build a Kafka message queue according to the preprocessing parameters.

[0024] Preferably, the process processing module further includes:

[0025] A data parsing unit is configured to parse the decrypted target multi-dimensional parameters to obtain parameters in JSON format, and send the parsed parameters to the data preprocessing unit.

[0026] An anomaly detection unit is configured to identify abnormal refrigerator states based on the preprocessing parameters, obtain anomaly parameters, and send the anomaly parameters to the task flow construction unit; the Kafka message queue is constructed based on the anomaly parameters.

[0027] Preferably, the parameter adjustment module is further configured to:

[0028] Read the corresponding parameter configuration; the parameter configuration is used to adjust the multi-dimensional parameters in the message queue.

[0029] Initialize the Bayesian optimization model to obtain the model initialization parameters;

[0030] A target probability model is constructed by training Gaussian process regression based on the model initialization parameters and the corresponding objective function.

[0031] Predict the target parameters of the refrigerator to be controlled based on the target probability model;

[0032] The target parameters are issued to obtain the state of the refrigerator to be controlled during the execution of the target parameters; when the preset state is not met, the target parameters are re-predicted, and the training data is updated using the target parameters from the previous round.

[0033] Preferably, the parameter adjustment module is further configured to:

[0034] When the refrigerator to be controlled is in the defrosting cycle, the target parameters will be issued in the next cycle.

[0035] Preferably, the control execution module includes:

[0036] A parameter sending unit, configured to send the target parameter to the refrigerator to be controlled;

[0037] An equipment operation status monitoring unit is configured to record the status of the refrigerator to be controlled during the execution of the target parameters;

[0038] An execution feedback unit is configured to feed back the state of the refrigerator to be adjusted during the execution of the target parameters to the parameter adjustment module.

[0039] Preferably, the system further includes:

[0040] An abnormality warning module is configured to issue an abnormality warning when the refrigerator to be controlled is in an abnormal state.

[0041] Preferably, the system further includes:

[0042] A visualization module, configured to visually display the control of the refrigerator to be controlled;

[0043] A storage module configured to store all data generated in the system according to timestamps.

[0044] As described above, this application provides a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization. The system includes a data acquisition module installed on the refrigerator to be controlled, configured to acquire target multi-dimensional parameters of the refrigerator in real time; a process processing module located in the cloud, configured to preprocess the target multi-dimensional parameters, construct a message queue based on the preprocessed parameters, and push the message queue; the multi-dimensional parameters in the message queue are the parameters of the refrigerator under abnormal conditions; a parameter adjustment module located in the cloud, configured to perform optimal parameter search on the multi-dimensional parameters in the message queue based on Bayesian optimization to obtain the target parameters; and a control execution module configured to adjust the operating parameters of the hardware devices in the refrigerator to be controlled according to the target parameters. This application solves the problems of low efficiency and inability to dynamically adjust refrigerator parameters in real time using the above system. Attached Figure Description

[0045] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to this application;

[0047] Figure 2 This is a schematic diagram of the data acquisition module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0048] Figure 3 This is a schematic diagram of the process processing module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0049] Figure 4 This is a schematic diagram of the control execution module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application. Detailed Implementation

[0050] 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, and 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.

[0051] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0052] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0053] Figure 1 This is a schematic diagram of a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0054] See Figure 1 As can be seen, this embodiment provides a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, the system comprising:

[0055] A data acquisition module is installed on the refrigerator to be controlled, and the data acquisition module is configured to acquire the target multi-dimensional parameters of the refrigerator to be controlled in real time.

[0056] Specifically, in this embodiment, the relevant parameters of the refrigerator to be controlled are obtained through the data acquisition module. Since the control of the refrigerator to be controlled in this technical solution is not a single parameter control, the obtained parameters are multi-dimensional.

[0057] Figure 2 This is a schematic diagram of the data acquisition module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0058] See Figure 2 Furthermore, in some embodiments, the data acquisition module includes:

[0059] A temperature monitoring unit, configured to acquire refrigerator equipment temperature parameters at several locations within the refrigerator to be regulated;

[0060] A power monitoring unit, configured to acquire compressor power parameters of the compressor in the refrigerator to be regulated;

[0061] An environmental monitoring unit is configured to acquire the ambient temperature and humidity parameters of the environment in which the refrigerator to be regulated is located, as well as the refrigerator door status parameters.

[0062] Specifically, in this embodiment, parameters of the corresponding types are obtained through the temperature monitoring unit, the power monitoring unit, and the environmental monitoring unit, respectively.

[0063] For example, IoT data is collected and encrypted via a Wi-Fi module before being reported. This includes device serial number (SN), temperature sensors in each compartment, ambient temperature, compressor status, door open / close status in each compartment, electric valve status, damper status, humidity, fan status, evaporator temperature status, defrosting status, etc.

[0064] Furthermore, in some embodiments, the data acquisition module further includes:

[0065] A data encryption unit is configured to encrypt the refrigerator device temperature parameters, the compressor power parameters, the ambient temperature parameters, the ambient humidity parameters, and the refrigerator door status parameters according to a first key to obtain the target multi-dimensional parameters.

[0066] A data sending unit is configured to report the target multi-dimensional parameters via the MQTT protocol.

[0067] Specifically, in this embodiment, since not all modules in this system are located on-site, the data transmission process needs to be encrypted to avoid data leakage and other problems. The data encryption unit uses the first key to encrypt the refrigerator equipment temperature parameters, the compressor power parameters, the ambient temperature parameters, the ambient humidity parameters, and the refrigerator door status parameters, and the data sending unit reports the target multi-dimensional parameters based on the MQTT protocol.

[0068] The system also includes:

[0069] The process processing module is located in the cloud and is configured to preprocess the target multi-dimensional parameters, construct a message queue based on the preprocessed multi-dimensional parameters, and push the message queue; the multi-dimensional parameters in the message queue are the parameters of the refrigerator to be controlled under abnormal conditions.

[0070] Specifically, in this embodiment, the process processing module constructs a message queue of the target multi-dimensional parameters in the cloud to achieve concurrent processing of massive amounts of data.

[0071] Figure 3 This is a schematic diagram of the process processing module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0072] See Figure 3It is understood that, furthermore, in some embodiments, the process processing module includes:

[0073] A data decryption unit, configured to decrypt the target multi-dimensional parameters based on a first public key;

[0074] A data preprocessing unit is configured to preprocess the decrypted target multi-dimensional parameters to obtain preprocessed parameters; the preprocessing includes outlier filtering and data standardization.

[0075] The task flow construction unit is built on a distributed stream processing platform based on Apache Flink and is configured to build a Kafka message queue according to the preprocessing parameters.

[0076] Specifically, in this embodiment, before constructing the message queue, the target multi-dimensional parameters are first decrypted using the first public key by the data decryption unit to obtain decrypted data. The data is then preprocessed by the data preprocessing unit to remove abnormal data. Finally, the message queue containing the corresponding parameters to be adjusted is constructed by the task flow construction unit.

[0077] Furthermore, in some embodiments, the process processing module further includes:

[0078] A data parsing unit is configured to parse the decrypted target multi-dimensional parameters to obtain parameters in JSON format, and send the parsed parameters to the data preprocessing unit.

[0079] An anomaly detection unit is configured to identify abnormal refrigerator states based on the preprocessing parameters, obtain anomaly parameters, and send the anomaly parameters to the task flow construction unit; the Kafka message queue is constructed based on the anomaly parameters.

[0080] Specifically, in this embodiment, since the data formats of the target multi-dimensional parameters are different, it is impossible to directly identify the data. Therefore, it is necessary to use the data parsing unit to parse the target multi-dimensional parameters to obtain parameters in the JSON format.

[0081] Furthermore, since parameters only need to be adjusted under abnormal conditions, including a large number of normal parameters in the message queue would greatly affect processing efficiency. Therefore, this embodiment also sets up the anomaly detection unit to select abnormal parameters and use only abnormal parameters as the content of the message queue.

[0082] For example, a distributed stream processing platform built on Apache Flink uses the Kafka message queue to handle data reception and buffering, supporting high-throughput data streams. Then comes the real-time data processing pipeline: data cleaning and normalization, feature engineering and extraction, state management and time-series analysis, and anomaly detection and alerting. This layer of architecture supports concurrent processing of massive amounts of data.

[0083] The system also includes:

[0084] A parameter adjustment module, located in the cloud, is configured to perform optimal parameter search on the multi-dimensional parameters in the message queue based on Bayesian optimization to obtain the target parameters.

[0085] Specifically, in this embodiment, the process of searching for the optimal parameters includes:

[0086] Read the corresponding parameter configuration; the parameter configuration is used to adjust the multi-dimensional parameters in the message queue.

[0087] Initialize the Bayesian optimization model to obtain the model initialization parameters;

[0088] A target probability model is constructed by training Gaussian process regression based on the model initialization parameters and the corresponding objective function.

[0089] Predict the target parameters of the refrigerator to be controlled based on the target probability model;

[0090] The target parameters are issued to obtain the state of the refrigerator to be controlled during the execution of the target parameters; when the preset state is not met, the target parameters are re-predicted, and the training data is updated using the target parameters from the previous round.

[0091] It should be noted that when the refrigerator to be controlled is in the defrosting cycle, the target parameters will be issued in the next cycle.

[0092] For example, the core of the Bayesian optimization algorithm is a Gaussian process regression model, aiming to improve the acquisition function. Bayesian optimization is a sequential optimization strategy based on Bayes' theorem, mainly used to solve black-box function optimization problems. In refrigerator parameter optimization, the relationship between the objective function (energy consumption, temperature stability, etc.) and the adjustment parameters is complex and unknown. Bayesian optimization finds the optimal parameter combination through continuous exploration and utilization.

[0093] The system also includes:

[0094] A control execution module is configured to adjust the operating parameters of the hardware devices in the refrigerator to be controlled according to the target parameters.

[0095] Specifically, in this embodiment, the control execution module adjusts the parameters.

[0096] Figure 4 This is a schematic diagram of the control execution module in a real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization, as described in this application.

[0097] See Figure 4 It can be seen that, further, in some embodiments, the control execution module includes:

[0098] A parameter sending unit, configured to send the target parameter to the refrigerator to be controlled;

[0099] An equipment operation status monitoring unit is configured to record the status of the refrigerator to be controlled during the execution of the target parameters;

[0100] An execution feedback unit is configured to feed back the state of the refrigerator to be adjusted during the execution of the target parameters to the parameter adjustment module.

[0101] Specifically, in this embodiment, the parameter distribution interface is called (it can also be distributed via Kafka messages, with the cloud and device using the MQTT protocol), device status is monitored, and execution effect feedback is provided (to understand whether the device parameter command has been executed and to add a retry mechanism).

[0102] Furthermore, in some embodiments, the system further includes:

[0103] An abnormality warning module is configured to issue an abnormality warning when the refrigerator to be controlled is in an abnormal state.

[0104] A visualization module, configured to visually display the control of the refrigerator to be controlled;

[0105] A storage module configured to store all data generated in the system according to timestamps.

[0106] Specifically, in this embodiment, the abnormality warning module provides early warnings for abnormal states, thereby achieving a dual management scheme of early warning and adjustment. The visualization module visualizes the entire process of system operation, allowing operators to monitor the system more intuitively. Finally, the storage module stores the data generated by the system, providing a data foundation for subsequent data traceability.

[0107] For example, the overall process of the system includes:

[0108] The refrigerator periodically reports its full-state data to the IoT platform via the MQTT protocol. The IoT platform parses the data into a universally recognizable JSON format and pushes the status messages to the real-time streaming parameter control center via Kafka. Immediate reporting is triggered when critical states change, such as compressor status, electric switching valve status, door open / close status, set temperature status of each compartment, and operating mode.

[0109] After receiving the device status message pushed by Kafka, the real-time streaming parameter control center first reads the parameter adjustment configuration D1 from Redis, which interacts with the accompanying parameter adjustment and monitoring system (monitoring and visualization layer). This includes information such as whether the current device SN needs automatic parameter adjustment, the specific parameter name, the corresponding set temperature, and the corresponding room integral average range (if not available, the default value is used).

[0110] Perform model initialization D2:

[0111] 1) Use Latin hypercube sampling to generate N sets (5 sets by default) of initial parameter combinations S1. It should be noted that the values ​​of different parameters are ranges. This range has been preset in the parameter tuning system software based on laboratory test results.

[0112] 2) The execution layer sequentially sends a set of predicted parameter values ​​S1 and S2. The timing of the sending is when the compressor stops at the end of this cooling cycle. If the monitored equipment does not respond, the corresponding retry mechanism is initiated.

[0113] 3) After the next compressor refrigeration cycle ends following the issuance of the command (configurable interval), the flow processing layer continues to collect data and calculates the integral average values ​​of the refrigerator compartment temperature, freezer compartment temperature, and variable temperature compartment temperature in real time. The formula for calculating the integral average value is as follows:

[0114]

[0115] The compressor running time is t1, the total compressor start-stop cycle time is t, and Δt i For the i-th temperature value, Temp i The duration is defined as follows: t1 is the start time of the calculation, t2 is the end time of the calculation, and n is the number of sensors in the compartment (S3). If defrosting occurs during this period, the previous command will be resent.

[0116] 4) Calculate the objective function J value S4. The overall objective function is designed as follows:

[0117]

[0118] Where α and β are weight values, α = 0.9 and β = 0.1 are set as default values, or they can be obtained through the D1 configuration. setTo set a temperature value, the objective function J is continuously adjusted to maximize the value of parameter C.

[0119] 5) After completing S4, update dataset S5 with the calculated J value and the issued parameter values.

[0120] 6) Check whether all N sets of initialized parameters have been issued and whether the objective function J value S6 has been calculated for each.

[0121] 7) Complete the initialization of the Bayesian optimizer. Use Gaussian process regression to build a probabilistic model of the parameters and the objective. If both are completed, the model initialization process ends (S7), and the initial GP model is constructed.

[0122] The parameters and corresponding J values ​​of the model initialization are updated to the Bayesian optimizer, and a probabilistic model D3 of the parameters and the objective is established through Gaussian process regression training.

[0123] After completing the Gaussian process regression training, the next evaluation point is selected as the parameter and the new predicted value D4 using the Expectation Improvement (EI) function.

[0124] Instructions are sent to device D5 through the control execution layer.

[0125] The compressor cooling cycle ends at D6. Check if defrosting occurred during this cycle; if so, resend the command.

[0126] If there was no defrosting in the previous compressor cooling cycle, calculate the integral mean of the temperature of each compartment during this cycle, the compressor running time, and the objective function J value D7.

[0127] Determine if the integral mean, compressor runtime, and J value all satisfy condition D8. If they do, record this data and update the dataset D9 to update the Gaussian process regression model D3.

[0128] Determine if the number of prediction iterations (the number of iterations is based on laboratory results and has a preset default value, or can be obtained from the D1 configuration) meets the set value D10. If not, return to D3 to continue with the next set of predictions.

[0129] If D10 satisfies the required number of iterations, results similar to those in Table 1 will be generated:

[0130] Table 1

[0131] Number of times C1 value C2 value J value 1 C11 C21 J1 2 C12 C22 J2 3 C13 C23 J3 4 C14 C24 J4 5 C15 C25 J5

[0132] The parameter set corresponding to the largest J value among all recorded predicted values ​​is used as the optimal parameter set for D11. A larger J value indicates a higher degree of fit.

[0133] After issuing the optimal parameters, wait for the compressor's cooling cycle to end, then continue calculating the integral mean, compressor runtime, and whether the J value meets condition D12. If defrosting occurs during this process, the optimal parameters will be reissued for the next cooling cycle. If the condition is not met, the iteration count is reset, and the prediction process repeats the D4 procedure.

[0134] If condition D12 is met, the compressor refrigeration cycle ends after waiting for 4 hours in D13.

[0135] Calculate the mean of the integral and the J value D14 for the time interval D13.

[0136] Determine whether the absolute value of the integral mean of D14 minus the set temperature meets the range of the room integral mean of D15.

[0137] If D15 does not meet the conditions, reset the iteration count and restart the prediction process, repeating the D4 procedure.

[0138] If D15 meets the conditions, the automatic parameter tuning ends.

[0139] The advantages of this embodiment include:

[0140] All operations during the entire parameter tuning process are logged in the monitoring and visualization layer for engineers to monitor and adjust later. The monitoring and visualization system can directly set the parameter tuning configuration and share the configuration with the real-time streaming parameter control center. It also controls the operation of the real-time streaming parameter control program. After parameter tuning is completed, it can be configured to continue iteration, at which point the device will re-enter the prediction process to continue iterating and fitting to find the optimal parameters.

[0141] Through the interaction between the real-time flow parameter control center and the visualization system, a large number of devices can be simultaneously adjusted, achieving a personalized experience for each device. The parameters of each device are optimized based on its actual performance under specific user conditions, allowing the refrigerator to achieve maximum efficiency.

[0142] Through the above steps, the entire system implements a method for real-time control of refrigerator parameters based on streaming Bayesian optimization. Each refrigerator is customized by constructing a probabilistic model of parameters and objectives using Gaussian process regression based on its own user operating conditions.

[0143] The classic Bayesian optimization algorithm was successfully applied to the real-time optimization of refrigerator parameters. Through streaming processing, parameter adjustment was achieved at the second level, which significantly improved the energy efficiency and temperature control accuracy of the refrigerator.

[0144] For ease of explanation, the above description has been provided in conjunction with specific embodiments. However, the discussion in some embodiments is not intended to be exhaustive or to limit the embodiments to the specific forms disclosed above. Various modifications and variations can be obtained based on the above teachings. The selection and description of the above embodiments are for the purpose of better explaining the contents of this disclosure, thereby enabling those skilled in the art to better utilize the embodiments.

Claims

1. A real-time cloud-based control system for refrigerator parameters based on streaming Bayesian optimization, characterized in that, The system includes: A data acquisition module is installed on the refrigerator to be controlled, and the data acquisition module is configured to acquire the target multi-dimensional parameters of the refrigerator to be controlled in real time. The process processing module is located in the cloud and is configured to preprocess the target multi-dimensional parameters, construct a message queue based on the preprocessed multi-dimensional parameters, and push the message queue; the multi-dimensional parameters in the message queue are the parameters of the refrigerator to be controlled in an abnormal state. A parameter adjustment module, which is located in the cloud, is configured to perform optimal parameter search on the multi-dimensional parameters in the message queue based on Bayesian optimization to obtain the target parameters. A control execution module is configured to adjust the operating parameters of the hardware devices in the refrigerator to be controlled according to the target parameters.

2. The real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 1, characterized in that, The data acquisition module includes: A temperature monitoring unit, configured to acquire refrigerator equipment temperature parameters at several locations within the refrigerator to be regulated; A power monitoring unit, configured to acquire compressor power parameters of the compressor in the refrigerator to be regulated; An environmental monitoring unit is configured to acquire the ambient temperature and humidity parameters of the environment in which the refrigerator to be regulated is located, as well as the refrigerator door status parameters.

3. The real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 2, characterized in that, The data acquisition module further includes: A data encryption unit is configured to encrypt the refrigerator device temperature parameters, the compressor power parameters, the ambient temperature parameters, the ambient humidity parameters, and the refrigerator door status parameters according to a first key to obtain the target multi-dimensional parameters. A data sending unit is configured to report the target multi-dimensional parameters via the MQTT protocol.

4. The real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 3, characterized in that, The process processing module includes: A data decryption unit, configured to decrypt the target multi-dimensional parameters based on a first public key; A data preprocessing unit is configured to preprocess the decrypted target multi-dimensional parameters to obtain preprocessed parameters; the preprocessing includes outlier filtering and data standardization. The task flow construction unit is built on a distributed stream processing platform based on Apache Flink and is configured to build a Kafka message queue according to the preprocessing parameters.

5. A real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 4, characterized in that, The process processing module also includes: A data parsing unit is configured to parse the decrypted target multi-dimensional parameters to obtain parameters in JSON format, and send the parsed parameters to the data preprocessing unit. An anomaly detection unit is configured to identify abnormal refrigerator states based on the preprocessing parameters, obtain anomaly parameters, and send the anomaly parameters to the task flow construction unit; the Kafka message queue is constructed based on the anomaly parameters.

6. The real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 1, characterized in that, The parameter adjustment module is also configured to: Read the corresponding parameter configuration; the parameter configuration is used to adjust the multi-dimensional parameters in the message queue. Initialize the Bayesian optimization model to obtain the model initialization parameters; A target probability model is constructed by training Gaussian process regression based on the model initialization parameters and the corresponding objective function. Predict the target parameters of the refrigerator to be controlled based on the target probability model; The target parameters are sent out, and the state of the refrigerator to be controlled during the execution of the target parameters is obtained; If the preset state is not met, the target parameters are re-predicted, and the training data is updated using the target parameters from the previous round.

7. A real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 6, characterized in that, The parameter adjustment module is also configured to: When the refrigerator to be controlled is in the defrosting cycle, the target parameters will be issued in the next cycle.

8. The real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 1, characterized in that, The control execution module includes: A parameter sending unit, configured to send the target parameter to the refrigerator to be controlled; An equipment operation status monitoring unit is configured to record the status of the refrigerator to be controlled during the execution of the target parameters; An execution feedback unit is configured to feed back the state of the refrigerator to be adjusted during the execution of the target parameters to the parameter adjustment module.

9. A real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 5, characterized in that, The system also includes: An abnormality warning module is configured to issue an abnormality warning when the refrigerator to be controlled is in an abnormal state.

10. A real-time cloud-based refrigerator parameter control system based on streaming Bayesian optimization according to claim 1, characterized in that, The system also includes: A visualization module, configured to visually display the control of the refrigerator to be controlled; A storage module configured to store all data generated in the system according to timestamps.

Images