An electric appliance intelligent temperature control method and system based on data analysis

Abstract

The application discloses an electric appliance intelligent temperature control method and system based on data analysis, and relates to the technical field of intelligent temperature control.The method comprises the following steps: collecting environment data, scanning space thermal field distribution, calculating initial thermal field uniformity index, and integrating into local data set through an edge controller; training a local thermodynamic model, generating an initial symbolic equation and calculating fitting error, obtaining equation coefficients and performance index; performing federal symbolic equation aggregation in the cloud, outputting a global optimization equation, and issuing to the edge controller to replace the local thermodynamic model and adjust the finite element grid heat conduction coefficient; constructing a digital twin environment in the edge controller, performing real-time thermal field regulation and control, and triggering local symbolic regression retraining according to the collected feedback.The application constructs a local thermodynamic model through data analysis, combines federal symbolic regression optimization and a digital twin environment, realizes high-precision and self-adaptive intelligent temperature control, and improves thermal field uniformity and temperature control energy efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent temperature control, and in particular to a method and system for intelligent temperature control of an electrical appliance based on data analysis. Background Art

[0002] With the development of intelligent control and Internet of Things technologies, the temperature control methods of electrical appliances are evolving from traditional mechanical or fixed rule control to data-driven intelligent regulation. Existing intelligent temperature control technologies mainly rely on feedback from ambient temperature sensors, combined with preset control logic or simple algorithms for adjustment. However, these methods are generally unable to accurately perceive the spatial thermal field distribution, resulting in limited temperature control accuracy and lack of adaptability to complex environmental changes. In addition, most existing controls are based on static models and fail to combine thermodynamic modeling and machine learning techniques to dynamically optimize temperature control strategies. Existing technologies have limited data collection and processing capabilities, making it difficult to accurately characterize the spatial thermal field distribution, resulting in local temperature unevenness; temperature control parameter adjustment relies on fixed rules and lacks optimization mechanisms such as symbolic regression and federated learning, resulting in insufficient adaptability; cloud-based collaborative optimization capabilities are weak, and distributed computing cannot be effectively utilized for real-time regulation. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides an intelligent temperature control method for electrical appliances based on data analysis to solve the problem of local temperature unevenness caused by the difficulty in accurately depicting the spatial thermal field distribution.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In the first aspect, the present invention provides an intelligent temperature control method for electrical appliances based on data analysis, which includes collecting environmental data, scanning the spatial thermal field distribution, calculating the initial thermal field uniformity index, and integrating it into a local data set through an edge controller; training a local thermodynamic model, generating an initial symbolic equation and calculating the fitting error, obtaining the equation coefficients and performance indicators, performing data enhancement, and uploading it to the cloud; performing federal symbolic equation aggregation in the cloud, outputting a global optimization equation, sending it to the edge controller, replacing the local thermodynamic model, and adjusting the finite element mesh thermal conductivity; building a digital twin environment in the edge controller, performing real-time thermal field control, triggering local symbolic regression retraining based on the collected feedback, and uploading the retrained coefficients to the cloud to participate in the next round of federal optimization.

[0007] As a preferred solution of the intelligent temperature control method for electrical appliances based on data analysis described in the present invention, the federated symbolic equation aggregation refers to selecting a non-dominated solution set through a Pareto front screening algorithm, performing a symbolic crossover operation and applying a Gaussian coefficient perturbation, and performing federated iterative evolution of the equation population.

[0008] As a preferred solution of the intelligent temperature control method for electrical appliances based on data analysis described in the present invention, the adjustment of the thermal conductivity of the finite element grid refers to dynamically adjusting the thermal conductivity of each grid according to the trend of change in the thermal field uniformity output by the global optimization equation, wherein the adjustment range of the thermal conductivity is proportional to the sensitivity of the thermal field uniformity to the local temperature, and the maximum adjustment range is controlled by a preset sensitivity coefficient.

[0009] As a preferred solution of the data analysis-based intelligent temperature control method for electrical appliances described in the present invention, the training of the local thermodynamic model refers to data cleaning and feature normalization of the local data set, using a polynomial regression algorithm to construct a local thermodynamic model, and training, verification and optimization through the local data set.

[0010] As a preferred solution of the intelligent temperature control method for electrical appliances based on data analysis described in the present invention, the local symbolic regression retraining means that when the deviation between the actual value of the thermal field uniformity calculated in real time and the predicted value of the local thermodynamic model exceeds the retraining threshold, Gaussian noise is added to the historical environmental data to generate an enhanced training set, the symbolic regression algorithm is re-run, and the retrained equation coefficients are obtained.

[0011] As a preferred solution of the intelligent temperature control method of an electrical appliance based on data analysis described in the present invention, the specific steps of constructing a digital twin environment and performing real-time thermal field control are as follows:

[0012] Define the state space, action space and reward function, and use the proximal policy optimization algorithm to train the policy network;

[0013] After training converges, the spatial thermal field distribution is scanned in real time, and the optimal action is selected based on the current state;

[0014] The stepper motor is driven to adjust the air outlet angle and the frequency converter is used to adjust the wind speed, and a pulse signal is sent to the piezoelectric material controller to change the local thermal conductivity.

[0015] As a preferred solution of the intelligent temperature control method for electrical appliances based on data analysis described in the present invention, the symbolic crossover operation refers to selecting two parent equations from the non-dominated solution set, exchanging their sub-expressions containing temperature difference terms and angle trigonometric function terms, generating a new equation, and applying random perturbations to the coefficients of the new equation.

[0016] In a second aspect, the present invention provides an intelligent temperature control system for electrical appliances based on data analysis, comprising a data integration module, a local model module, a cloud aggregation module, and a real-time control module;

[0017] The data integration module is used to collect environmental data, scan the spatial thermal field distribution, calculate the initial thermal field uniformity index, and integrate it into a local data set through the edge controller;

[0018] The local model module is used to train the local thermodynamic model, generate the initial symbolic equation and calculate the fitting error, obtain the equation coefficients and performance indicators, perform data enhancement, and upload it to the cloud;

[0019] The cloud aggregation module is used to aggregate federated symbolic equations in the cloud, output global optimization equations, send them to the edge controller, replace the local thermodynamic model, and adjust the thermal conductivity of the finite element mesh;

[0020] The real-time control module is used to build a digital twin environment on the edge controller, perform real-time thermal field control, trigger local symbolic regression retraining based on collected feedback, and upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

[0021] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for intelligent temperature control of electrical appliances based on data analysis as described in the first aspect of the present invention is implemented.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for intelligent temperature control of electrical appliances based on data analysis as described in the first aspect of the present invention is implemented.

[0023] The beneficial effects of the present invention are as follows: by collecting environmental data and constructing a local data set, accurate perception of the spatial thermal field is achieved, providing data support for subsequent thermodynamic modeling; by training the local thermodynamic model, generating symbolic equations and optimizing the fitting error, the temperature control strategy is made more adaptable, and the optimized model is uploaded to the cloud; federal symbolic equations are aggregated in the cloud to generate global optimization equations and used to dynamically adjust the thermal conductivity of the finite element mesh, thereby improving the uniformity of the thermal field and the temperature control accuracy; a digital twin environment is constructed on the edge controller, local symbolic regression retraining is triggered based on real-time data feedback, and optimized parameters are uploaded to continuously iterate the global model, thereby achieving accurate, efficient and adaptive intelligent temperature control, improving energy efficiency and optimizing user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is an architecture diagram of an intelligent temperature control system for electrical appliances based on data analysis in Example 1.

[0026] Figure 2 This is a flow chart of an intelligent temperature control method for an electrical appliance based on data analysis in Example 1.

[0027] Figure 3 Schematic diagram of federated symbolic equation aggregation in Example 1.

[0028] Figure 4 Schematic diagram of digital twin control in Example 1. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0030] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0031] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0032] Example 1, reference Figures 1 to 4 , which is the first embodiment of the present invention, provides an intelligent temperature control method for an electrical appliance based on data analysis, comprising the following steps:

[0033] S1: Collect environmental data, scan the spatial thermal field distribution, calculate the initial thermal field uniformity index, and integrate it into a local data set through the edge controller.

[0034] Specifically, the following steps are included:

[0035] S1.1: Deploy the sensor array.

[0036] Specifically, an infrared thermal imaging sensor array is deployed in the target space (such as a room or equipment cabin), and the sensor spacing is distributed in a 0.5m×0.5m grid (one layer is deployed every 1m in the vertical direction) to ensure three-dimensional space coverage.

[0037] Each grid point is equipped with a multimodal sensor node, integrating:

[0038] Temperature sensor (accuracy ±0.2°C, range -20°C to 80°C).

[0039] Humidity sensor (accuracy ±3%RH, range 0~100%RH).

[0040] Wind speed sensor (ultrasonic, range 0~10m / s, resolution 0.1m / s).

[0041] The multimodal sensor nodes communicate with the edge controller via the Zigbee 3.0 wireless protocol with a transmission period of 100ms and support for dynamic adjustment of the sampling frequency (±20ms).

[0042] During initial deployment, cross-calibration was performed in a constant temperature and humidity chamber (25°C ± 0.1°C, 50% RH ± 1%). A five-point calibration method (0%, 25%, 50%, 75%, and 100% of the range) was applied to each sensor node, and the output curve was recorded. Calibration was performed using the least squares method, and after calibration, sensor errors were controlled within ±0.1°C (temperature), ±2% RH (humidity), and ±0.05 m / s (wind speed).

[0043] S1.2: Collect environmental data.

[0044] Specifically, the IEEE 1588 PTP precise time protocol is used to achieve data synchronization. The master clock is synchronized with the clock of each sensor node, and the time deviation is ≤1ms.

[0045] A four-tuple identifier is attached to each data packet, including spatial coordinates, sensor ID, timestamp, and data type.

[0046] The collected environmental data include temperature, humidity and wind speed.

[0047] The MQTT protocol is used for data transmission, the QoS level is set to 1 (at least once delivery), and the bandwidth usage is ≤10%.

[0048] S1.3: Calculation of initial thermal field uniformity index.

[0049] Specifically, Gaussian filtering is performed on the temperature data to eliminate sensor noise, and the spatial average temperature is calculated based on the temperature data.

[0050] Calculate the initial thermal field uniformity index, the expression is:

[0051] ;

[0052] in, is the initial thermal field uniformity index, is the location weight (the weight of the device surrounding area is set to 2.0, and the general area is 1.0), is the spatial coordinate Temperature data at is the average temperature of the space, is the maximum temperature of the space at the current moment, is the minimum temperature of the space at the current moment, is the three-dimensional coordinate in space.

[0053] S1.4: Perform abnormal environment data processing.

[0054] It should be noted that a sliding window is used to calculate the temperature change rate for mutation detection. If the temperature change rate is greater than 2°C / s, it is marked as a mutation anomaly.

[0055] Data outliers are determined based on the Mahalanobis distance. If the Mahalanobis distance is greater than 3.0, it is determined to be a sensor failure and the node self-test is triggered.

[0056] For continuous missing data (within 3 seconds), the spatiotemporal Kriging interpolation method is used to fill the missing values. If it is greater than three seconds, the sensor is judged to be faulty and the node self-check is triggered.

[0057] For data marked as mutation anomalies, exponentially weighted moving average is used for smoothing and replacing outliers.

[0058] S1.5: Local dataset integration.

[0059] It should be understood that the processed data is formatted and saved in JSON format. Data compression is performed using the LZ77 algorithm and encryption is performed using AES-256-GCM mode. The resulting local dataset is stored on the edge node, retaining the last 30 days of historical data.

[0060] Optimally, high-quality environmental data collection is achieved through precise sensor array layout, high-precision calibration, and multimodal data fusion. IEEE 1588 PTP protocol is used to ensure clock synchronization, and MQTT protocol is used to optimize data transmission, improving data consistency and transmission efficiency. Gaussian filtering for noise reduction, mutation detection, Mahalanobis distance outlier identification, and missing value filling techniques are used to improve data collection accuracy and reliability. Furthermore, the thermal field uniformity index is used to quantify spatial temperature distribution, providing data support for thermal management optimization. Data storage uses JSON formatting, LZ77 compression, and AES-256-GCM encryption to ensure efficient and secure storage. Historical data from the last 30 days is retained to facilitate trend analysis and fault tracing.

[0061] S2: Train the local thermodynamic model, generate the initial symbolic equation and calculate the fitting error, obtain the equation coefficients and performance indicators, perform data enhancement, and upload to the cloud.

[0062] Specifically, the following steps are included:

[0063] S2.1: Preprocess the local dataset.

[0064] Specifically, environmental features are extracted from the local data set, including calculating the temperature field gradient through the central difference method, and calculating the humidity change rate and wind speed direction angle.

[0065] Min-Max normalization is used to map the extracted environmental features to the [0, 1] interval, and the temperature field gradient is logarithmically transformed to suppress dimensional differences.

[0066] S2.2: Construct a local thermodynamic model using a polynomial regression algorithm.

[0067] Specifically, a third-order polynomial regression is used to construct the local thermodynamic model, which is expressed as follows:

[0068] ;

[0069] in, Predict output for local thermodynamic models, is the intercept term, is the coefficient of the first-order term, is the interaction coefficient, is the third-order coefficient, is the input environmental characteristic variable, is another input environment characteristic variable, is the number of environmental features, It is the index of the number of environmental features, and its value range is , is greater than or equal to Another environmental feature quantity index, the value range is .

[0070] L2 regularization is added to the loss function, and the L-BFGS algorithm is used to solve the optimal coefficient. The maximum number of iterations is set to 500, and the convergence threshold is 1e-6.

[0071] S2.3: Calculate the fitting error.

[0072] It should be noted that the fitting error includes the mean absolute error (MAE), the root mean square error (RMSE) and the coefficient of determination) ( ).

[0073] The model was divided into training set (70%), validation set (15%) and test set (15%). If the RMSE of the validation set was greater than 1.2 × the RMSE of the training set, it was considered as overfitting and the polynomial order was reduced and retrained.

[0074] S2.4: Obtain equation coefficients and performance indicators.

[0075] Specifically, the coefficient matrix is ​​extracted from the local thermodynamic model and the intercept term , stored in JSON format in the order of environmental characteristics.

[0076] At the same time, the index is filtered and only Local thermodynamic model parameters with a MAE < 0.85 and a MAE < 0.05, with coefficients having a precision of 4 decimal places.

[0077] S2.5: Perform data augmentation on the local dataset.

[0078] It should be noted that data augmentation refers to adding anisotropic Gaussian noise to each sample in the local dataset to generate an enhanced dataset that is 10 times the size of the original data.

[0079] S2.6: Package the equation coefficients, performance metrics, and augmented datasets and upload them to the cloud.

[0080] Specifically, the equation coefficients, performance indicators, and enhanced datasets are packaged into the Protobuf binary format, which includes: local thermodynamic model metadata (training time, edge node ID, hardware version), equation coefficient matrix, performance indicators, and enhanced dataset.

[0081] Data is encrypted using TLS 1.3, with certificates issued by a cloud-based CA and a SHA-256 hash checksum appended to the data packet. The transmission frequency is set to once every six hours, with a peak bandwidth limit of 50Mbps.

[0082] Optimally, the quality of the local dataset is improved through feature extraction, normalization, and regularization. A local thermodynamic model is trained using third-order polynomial regression combined with the L-BFGS optimization algorithm, with the order dynamically adjusted to prevent overfitting and ensure high accuracy and generalization. Anisotropic Gaussian noise is used for data augmentation to improve the adaptability of the local thermodynamic model. Data is stored in Protobuf format and transmitted using TLS 1.3 encryption with a SHA-256 hash checksum to ensure efficiency and security.

[0083] S3: Federated symbolic equation aggregation is performed in the cloud, global optimization equations are output, and the equations are sent to the edge controller to replace the local thermodynamic model and adjust the thermal conductivity of the finite element mesh.

[0084] Specifically, the following steps are included:

[0085] S3.1: Select the non-dominated solution set through the Pareto front screening algorithm.

[0086] Specifically, the input is a set of local equation coefficients uploaded by each edge node, and each equation coefficient set is associated with a performance indicator.

[0087] The optimization goal is to maximize , minimize MAE and minimize equation complexity (number of terms).

[0088] The dominance relationship is determined as follows: Equation 1 dominates Equation 2 if and only if:

[0089] ;

[0090] And at least one of the objectives is strictly better.

[0091] Furthermore, the NSGA-II algorithm is specifically implemented as follows: fast non-dominated sorting, stratifying all equations according to dominance relationships to generate non-dominated hierarchies; calculating congestion, sorting equations at the same level according to target space distance to preserve diversity; and selecting the top 50% of equations based on the congestion screening results to form a non-dominated solution set.

[0092] S3.2: Perform symbol-level crossover operation.

[0093] Specifically, two parent equations are selected from the non-dominated solution set, and their sub-expressions containing temperature difference terms and angle trigonometric function terms are exchanged to generate new child equations.

[0094] At the same time, invalid cross-results are eliminated, such as items that have no physical meaning. .

[0095] S3.3: Apply Gaussian coefficient perturbation, perform equation population federation iteration, and output the global optimization equation.

[0096] Specifically, an adaptive Gaussian perturbation is applied to the coefficients of the child equations, and the expression is:

[0097] ;

[0098] in, are the coefficients of the sub-equation after perturbation, are the coefficients of the original sub-equation, is the Gaussian noise term, is a Gaussian distribution, is the adaptively adjusted noise variance.

[0099] If after the disturbance If it is increased, the disturbance intensity will be amplified: .

[0100] The termination condition of the equation population federation iteration is set to reach the maximum number of iterations (50 times) or the global equation is on the validation set. Improvement of <0.1% for 5 consecutive times.

[0101] The top 10% of the best equations are retained in each round to avoid performance degradation. Finally, the Pareto frontier is selected. The highest equation with a complexity of ≤5 terms is regarded as the global equation.

[0102] S3.4: Send the global optimization equation to the edge controller to replace the local thermodynamic model.

[0103] Specifically, the global optimization equation is encoded into JSON-LD format and sent to the edge controller via the HTTP / 2 protocol. The data packet is signed with an HMAC-SHA256 signature to prevent tampering.

[0104] The edge controller loads the global optimization equation into the memory buffer and verifies ≥0.85. Atomically replaces the running local thermodynamic model. The old version is retained for 24 hours for rollback.

[0105] S3.5: Adjust the finite element mesh thermal conductivity.

[0106] Specifically, the sensitivity of each grid point temperature to the initial thermal field uniformity index is calculated according to the global optimization equation, and the expression is:

[0107] ;

[0108] in, is the grid point Thermal field uniformity sensitivity at is the partial derivative of thermal field uniformity with respect to grid point temperature, is the thermal field uniformity versus environmental characteristic variables The partial derivative of Environmental characteristic variables Temperature The partial derivative of .

[0109] The dynamic adjustment expression of thermal conductivity is:

[0110] ;

[0111] in, is the updated grid point The thermal conductivity, is the current thermal conductivity, is the sensitivity coefficient ( ), control the adjustment range of thermal conductivity to prevent instability caused by excessive changes, The maximum thermal conductivity is set to limit the growth range of the thermal conductivity to prevent the value from being too large and affecting the calculation stability.

[0112] The thermal conductivity is updated in real time through the API of finite element analysis software (such as COMSOL), and the heat conduction equation is solved iteratively until convergence (residual < 1e-4).

[0113] Optimally, thermodynamic models are optimized through federated symbolic equation aggregation to improve temperature control performance. Pareto front screening is used to select equation sets with high precision, low error, and low complexity, and symbolic crosstalk is used to enhance generalization capabilities. In Gaussian coefficient perturbation optimization, equation coefficients are adaptively adjusted to improve fit and prevent model degradation. The optimized global equations are securely delivered to the edge controller in JSON-LD format via HTTP / 2+HMAC-SHA256, replacing the local model and providing a rollback mechanism to ensure stability. Based on the global equations, the thermal conductivity of the finite element mesh is optimized, the thermal field uniformity sensitivity is calculated, and the thermal conductivity is dynamically adjusted through the COMSOL API to accelerate thermal balance and improve energy efficiency.

[0114] S4: Build a digital twin environment on the edge controller to perform real-time thermal field control, trigger local symbolic regression retraining based on collected feedback, and upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

[0115] Specifically, the following steps are included:

[0116] S4.1: Build a digital twin environment on the edge controller.

[0117] Specifically, a three-dimensional mesh model was generated using the Open3D framework, and the mesh resolution was consistent with the physical space (0.5m×0.5m×1m).

[0118] The updated finite element mesh thermal conductivity is mapped to the digital twin model. Data synchronization between the physical space and the digital twin is achieved through the OPC UA protocol. The temperature field data is updated every 100 milliseconds, and the wind speed and humidity data are updated every 500 milliseconds. The mean square error (MSE) between the digital twin's predicted temperature and the actual measured value is calculated, with an MSE requirement of less than 0.1°C.

[0119] S4.2: Update the policy network through the PPO algorithm.

[0120] Specifically, the state space is defined, including temperature, temperature field gradient, wind speed, temperature change rate, and real-time power of the device (collected from the device interface).

[0121] Define the action space, including temperature adjustment, wind speed adjustment, and finite element mesh thermal conductivity adjustment.

[0122] Furthermore, the architecture of the PPO algorithm policy network is as follows: input layer: 8-dimensional state → hidden layer: 256 neurons (ReLU) → output layer: 3-dimensional action mean and logarithmic standard deviation.

[0123] The discount factor is set to 0.99, the GAE parameter is set to 0.95, the strategy update step size is set to 0.2, and the learning rate is set to .

[0124] In each round, 2000 steps of interaction data are collected, the mini-batch size is 64, and 3 epochs are iterated. Convergence is determined when the average reward fluctuation is less than 1% for 10 consecutive rounds.

[0125] S4.3: Generate control signals based on the strategy network to perform real-time thermal field control.

[0126] Specifically, the policy network outputs the action probability distribution and samples it to obtain control instructions. The stepper motor is driven to adjust the air outlet angle, requiring a resolution of 0.5° and a response time of ≤50ms.

[0127] The fan speed is controlled by a frequency converter with an accuracy of ±0.05m / s.

[0128] Adjust the local thermal conductivity, driving voltage of piezoelectric material, and linear relationship between voltage and thermal conductivity.

[0129] After each action is executed, check If the rate of change is greater than 0.02 / s, it is considered effective. If the control is invalid for three consecutive times, the maximum wind speed is forced to perform emergency cooling.

[0130] S4.4: Collect feedback based on the control results, trigger local symbol regression retraining, and generate retraining coefficients.

[0131] It should be understood that retraining is initiated when the following conditions are met:

[0132] ;

[0133] in, To trigger the number of sampling points within the judgment window, is the sampling point index, To predict the thermal field uniformity, is the upper limit of the forecast deviation.

[0134] Noise injection is performed on historical environmental data, and the expression is:

[0135] ;

[0136] in, is the enhanced temperature data, used to expand the training set, is the original temperature data, is the noise amplitude coefficient, which is set to 0.5 to avoid excessive distortion. For space Axial length, is a Gaussian distributed random noise with a mean of 0 and a standard deviation of 0.2.

[0137] Retraining is performed using a deep symbolic regression network, and the loss function is added with an equation complexity penalty, expressed as:

[0138] ;

[0139] in, is the total loss function, is the mean square error, is the number of mathematical terms in the symbolic equation.

[0140] After training, new coefficients are obtained, requiring .

[0141] S4.5: Upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

[0142] Preferably, real-time thermal field control can be achieved through a digital twin environment to improve regulation accuracy. A physical-virtual synchronization mechanism is established based on Open3D and OPCUA, and the prediction error is constrained. PPO reinforcement learning is used to optimize the control strategy, adjust the temperature, wind speed and thermal conductivity, and control the thermal field with the optimal strategy to ensure 0.5° angle accuracy, 50ms response time and ±0.05m / s wind speed control. If the prediction deviation of the thermal field uniformity exceeds the threshold, local symbolic regression retraining is triggered, the training set is expanded through noise enhancement, and the symbolic equation is optimized. The retraining coefficients are uploaded to the cloud to participate in the next round of federated optimization, realize edge intelligent self-evolution, and improve the adaptability and stability of thermal field control.

[0143] This embodiment also provides an intelligent temperature control system for electrical appliances based on data analysis, comprising: a data integration module, a local model module, a cloud aggregation module, and a real-time control module;

[0144] The data integration module is used to collect environmental data, scan the spatial thermal field distribution, calculate the initial thermal field uniformity index, and integrate it into a local data set through the edge controller;

[0145] The local model module is used to train the local thermodynamic model, generate the initial symbolic equation and calculate the fitting error, obtain the equation coefficients and performance indicators, perform data enhancement, and upload it to the cloud;

[0146] The cloud aggregation module is used to aggregate federated symbolic equations in the cloud, output global optimization equations, and send them to the edge controller to replace the local thermodynamic model and adjust the thermal conductivity of the finite element mesh;

[0147] The real-time control module is used to build a digital twin environment on the edge controller, perform real-time thermal field control, trigger local symbolic regression retraining based on collected feedback, and upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

[0148] This embodiment also provides a computer device, which is suitable for the case of an intelligent temperature control method for electrical appliances based on data analysis, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent temperature control method for electrical appliances based on data analysis proposed in the above embodiment.

[0149] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0150] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for realizing intelligent temperature control of an electrical appliance based on data analysis as proposed in the above embodiment is implemented. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0151] In summary, the present invention achieves accurate perception of the spatial thermal field by: collecting environmental data and constructing a local data set, providing data support for subsequent thermodynamic modeling; by training the local thermodynamic model, generating symbolic equations and optimizing the fitting error, so that the temperature control strategy has stronger adaptability, and uploading the optimized model to the cloud; performing federal symbolic equation aggregation in the cloud, generating a global optimization equation and using it to dynamically adjust the thermal conductivity of the finite element mesh, thereby improving the thermal field uniformity and temperature control accuracy; building a digital twin environment on the edge controller, triggering local symbolic regression retraining based on real-time data feedback, and uploading optimized parameters to continuously iterate the global model, thereby achieving accurate, efficient and adaptive intelligent temperature control, improving energy efficiency and optimizing user experience.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent temperature control method for electrical appliances based on data analysis, characterized in that: include, Collect environmental data, scan the spatial thermal field distribution, calculate the initial thermal field uniformity index, and integrate it into a local data set through the edge controller; Train the local thermodynamic model, generate the initial symbolic equations and calculate the fitting error, obtain the equation coefficients and performance indicators, perform data enhancement, and upload to the cloud; The specific steps of training the local thermodynamic model are as follows: The local thermodynamic model is constructed using third-order polynomial regression, and the expression is: ； in, Predict output for local thermodynamic models, is the intercept term, is the coefficient of the first-order term, is the interaction coefficient, is the third-order coefficient, is the input environmental characteristic variable, is another input environment characteristic variable, is the number of environmental features, It is the index of the number of environmental features, and its value range is , is greater than or equal to Another environmental feature quantity index, the value range is ; Federated symbolic equations are aggregated in the cloud, global optimization equations are output, and sent to the edge controller to replace the local thermodynamic model and adjust the thermal conductivity of the finite element mesh. The specific steps of adjusting the thermal conductivity of the finite element grid are as follows: The sensitivity of each grid point temperature to the initial thermal field uniformity index is calculated according to the global optimization equation. The expression is: ； in, is the grid point Thermal field uniformity sensitivity at is the partial derivative of thermal field uniformity with respect to grid point temperature, is the thermal field uniformity versus environmental characteristic variables The partial derivative of Environmental characteristic variables Temperature The partial derivative of The dynamic adjustment expression of thermal conductivity is: ； in, is the updated grid point The thermal conductivity, is the current thermal conductivity, is the sensitivity coefficient ( ), control the adjustment range of thermal conductivity to prevent instability caused by excessive changes, The maximum thermal conductivity is set to limit the growth range of the thermal conductivity to prevent the value from being too large and affecting the calculation stability; Build a digital twin environment on the edge controller to perform real-time thermal field control, trigger local symbolic regression retraining based on collected feedback, and upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

2. The method for intelligent temperature control of an electrical appliance based on data analysis according to claim 1, characterized in that: The federated symbolic equation aggregation refers to selecting a non-dominated solution set through a Pareto front screening algorithm, performing a symbolic crossover operation and applying a Gaussian coefficient perturbation, and performing federated iterative evolution of the equation population.

3. The method for intelligent temperature control of an electrical appliance based on data analysis according to claim 1, characterized in that: The training of the local thermodynamic model refers to performing data cleaning and feature normalization on the local data set, building the local thermodynamic model using a polynomial regression algorithm, and performing training, verification, and optimization using the local data set.

4. The method for intelligent temperature control of an electrical appliance based on data analysis according to claim 1, characterized in that: The local symbolic regression retraining means that when the deviation between the actual value of the thermal field uniformity calculated in real time and the predicted value of the local thermodynamic model exceeds the retraining threshold, Gaussian noise is added to the historical environmental data to generate an enhanced training set, and the symbolic regression algorithm is re-run to obtain the retrained equation coefficients.

5. The method for intelligent temperature control of an electrical appliance based on data analysis according to claim 1, characterized in that: The digital twin environment is constructed to perform real-time thermal field control. The specific steps are: Define the state space, action space and reward function, and use the proximal policy optimization algorithm to train the policy network; After training converges, the spatial thermal field distribution is scanned in real time, and the optimal action is selected based on the current state; The stepper motor is driven to adjust the air outlet angle and the frequency converter is used to adjust the wind speed, and a pulse signal is sent to the piezoelectric material controller to change the local thermal conductivity.

6. The method for intelligent temperature control of an electrical appliance based on data analysis according to claim 2, characterized in that: The symbolic crossover operation refers to selecting two parent equations from the non-dominated solution set, exchanging their sub-expressions containing temperature difference terms and angle trigonometric function terms, generating a new equation, and applying random perturbations to the coefficients of the new equation.

7. An intelligent temperature control system for an electrical appliance based on data analysis, based on the intelligent temperature control method for an electrical appliance based on data analysis according to any one of claims 1 to 6, characterized in that: Including data integration module, local model module, cloud aggregation module and real-time control module; The data integration module is used to collect environmental data, scan the spatial thermal field distribution, calculate the initial thermal field uniformity index, and integrate it into a local data set through the edge controller; The local model module is used to train the local thermodynamic model, generate the initial symbolic equation and calculate the fitting error, obtain the equation coefficients and performance indicators, perform data enhancement, and upload it to the cloud; The cloud aggregation module is used to aggregate federated symbolic equations in the cloud, output global optimization equations, send them to the edge controller, replace the local thermodynamic model, and adjust the thermal conductivity of the finite element mesh; The real-time control module is used to build a digital twin environment on the edge controller, perform real-time thermal field control, trigger local symbolic regression retraining based on collected feedback, and upload the retraining coefficients to the cloud to participate in the next round of federated optimization.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the electrical appliance intelligent temperature control method based on data analysis according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the electrical appliance intelligent temperature control method based on data analysis according to any one of claims 1 to 6 are implemented.

Images