Cooperative operation method of smart home system

Abstract

The invention relates to the technical field of smart home cooperative control, and discloses a cooperative operation method of a smart home system, and the method comprises the steps: firstly constructing a smart home equipment state prediction model and an equipment cooperative control decision model, and predicting the operation state of each piece of intelligent equipment in the future through collecting user habit data and home environment data; and generating a preliminary collaborative operation instruction. Thirdly, a difference fusion model is constructed, the model optimizes the preliminary instruction according to the difference between the predicted operation state and the actual operation state of the equipment, and an optimal cooperative operation instruction is generated; and the system continuously monitors the actual operation state of the equipment, compares the actual operation state with a prediction result, and immediately triggers the difference fusion model to recalculate an instruction once a significant deviation is found, thereby ensuring that the smart home system is always kept in an optimal operation state. Through combination of prediction and real-time optimization, the cooperation efficiency and response speed of the smart home system are improved, and more comfortable and convenient smart home experience is provided for users.

Description

Technical Field

[0001] The present invention relates to the technical field of smart home collaborative control, and in particular to a collaborative operation method of a smart home system. Background Art

[0002] With the rapid development of science and technology, smart home systems have become an important part of modern family life. These systems provide residents with a more convenient, comfortable and safe living environment by integrating various smart devices, such as smart lighting, smart temperature control, smart security, etc. However, to maximize the benefits of smart home systems, not only do the various smart devices need to operate efficiently, but they also need to work together to adapt to the actual needs of users and changes in the home environment.

[0003] In the existing smart home systems, although each smart device has a certain degree of automation control capabilities, there are still many challenges in the collaborative operation between devices. On the one hand, because smart devices of different brands and types may use different communication protocols and interface standards, it becomes difficult to share information and interconnect them. On the other hand, even if some smart home systems can support collaborative work between devices, due to the lack of intelligent decision-making support, they can only achieve simple linkage control and cannot make adaptive adjustments according to the actual needs of users and changes in the home environment. In order to solve the above problems, in recent years, some researchers have begun to try to use artificial intelligence and machine learning technologies to build collaborative operation models of smart home systems. These models can automatically adjust and optimize the collaborative operation strategies of devices by real-time monitoring and analysis of the operating status of smart home devices, user habit data, and home environment data, thereby providing users with a more personalized service experience. However, the existing collaborative operation methods of smart home systems still have some shortcomings, such as low model training efficiency and inaccurate collaborative operation strategies. Summary of the invention

[0004] The purpose of the present invention is to provide a collaborative operation method of a smart home system to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for collaborative operation of a smart home system, the method comprising: Step A: Construct a smart home device state prediction model and a device collaborative control decision model; The smart home device state prediction model is used to predict the operating state of each smart device in the home within a period of time in the future, and its input data includes: User habit data, including user behavior patterns, preference settings, and historical operation records in smart homes; Home environment data, including temperature, humidity, light intensity, occupancy distribution and time information within the smart home; The output data are: the future operating status of each smart device, including the predicted switch status, predicted working mode, predicted working parameters and predicted working time of the device; The device collaborative control decision model is used to control the collaborative work of each smart device; its input data is: the predicted future operating state of each smart device output by the smart home device state prediction model; its output data is: the preliminary collaborative operation instructions of each smart device; Step B: constructing a difference fusion model, which is used to optimize the preliminary collaborative operation instructions; its input data is: the difference between the predicted operation state output by the smart home device state prediction model and the actual operation state of the device obtained by real-time monitoring of the smart home system, including device state difference, working time difference and working mode difference; its output data is: the optimal collaborative operation instruction of the smart home system after optimizing the preliminary collaborative operation instruction; Step C: Based on the smart home device state prediction model and the device collaborative control decision model, combined with actual user habit data and actual home environment data, generate preliminary collaborative operation instructions for each smart device to control the collaborative operation of each smart device, continuously monitor the actual operating status of each smart device, and compare it with the prediction results of the smart home device state prediction model. If significant deviations are found, the difference fusion model is immediately triggered to recalculate the optimal collaborative operation instructions.

[0006] Preferably, in step A, a convolutional neural network (CNN) algorithm is used to construct a smart home device state prediction model.

[0007] Preferably, the training step of the smart home device state prediction model includes: Collect training data, including user habit data, home environment data and corresponding actual operation status data of each smart device; user habit data includes user behavior patterns, preference settings and historical operation records in the smart home; home environment data includes temperature, humidity, light intensity, personnel distribution and time information in the smart home; actual operation status data of the device includes the switch status, working mode, working parameters and actual working time of the device; Preprocess the collected training data, including data cleaning and data standardization; Use the convolutional neural network (CNN) algorithm to build the initial structure of the smart home device status prediction model and set the model's hyperparameters; The initial model is trained using the preprocessed data set, and the model parameters are adjusted through an iterative optimization algorithm so that the model can accurately predict the operating status of each smart device in the home in the future. The labels of the training data are the actual operating status data of each smart device. During the training process, the cross-validation method was used to evaluate the model to prevent overfitting, and the model with the best performance was selected as the final smart home device status prediction model.

[0008] Preferably, the data cleaning method includes: For missing values, linear interpolation algorithm is used to fill them; For outliers, the threshold judgment method is used to regard data points that exceed the preset threshold range as outliers and replace them with the mean of their neighboring points.

[0009] Preferably, the method for data standardization includes: using the Z-score standardization method to standardize the cleaned data, specifically: subtracting the mean of each numerical data and dividing it by its standard deviation, so that the processed data conforms to the standard normal distribution, and the formula for standardization is: ,in, is the original data, is the mean of the original data, is the standard deviation of the original data, is the standardized data.

[0010] Preferably, in step A, a deep deterministic policy gradient (DDPG) algorithm is used to construct a device collaborative control decision model.

[0011] Preferably, the training step of the equipment collaborative control decision model includes: Step 1: Construct the network structure of the deep deterministic policy gradient DDPG algorithm, including the actor network and the critic network; the actor network is responsible for generating preliminary collaborative operation instructions for each intelligent device to represent the action; the critic network is responsible for evaluating the value of the action generated by the actor network; Step 2: Collect training data, including the predicted future operating state of each smart device output by the smart home device state prediction model as state input, and the corresponding optimal collaborative operation instructions annotated by humans as labels to represent the target action; Step 3: Initialize the parameters of the actor network and the critic network, and set the hyperparameters of the DDPG algorithm; Step 4: During the training process, the actor network generates an action based on the current state, the critic network evaluates the value of the action, and updates the parameters of the actor network and the critic network through the gradient descent method, so that the actor network generates actions that are closer to the optimal collaborative operation instructions; Step 5: Repeat step 4 until the performance index of the model reaches the preset standard, and a trained equipment collaborative control decision model is obtained.

[0012] Preferably, in step B, a random forest RF algorithm is used to construct a difference fusion model.

[0013] Preferably, the training step of the difference fusion model includes: Collect training data, including the difference between the predicted operating state output by the smart home device state prediction model and the actual operating state of the device obtained by real-time monitoring of the smart home system, including differences in device state, working time, and working mode; at the same time, collect manually annotated optimal collaborative operation instructions under the given difference between the predicted operating state and the actual operating state as labels for the training data; Build a random forest RF algorithm model and set the parameters of the random forest, including the number of trees, maximum depth, and minimum number of sample splits; Use the training data to train the random forest RF algorithm model, and through iterative learning, enable the model to accurately predict the optimal collaborative operation instructions under given input differences; After the training is completed, the difference fusion model is obtained.

[0014] Preferably, the method for determining whether the deviation between the actual operating state and the predicted operating state of the smart home device is significant is: setting a difference threshold, comparing the calculated difference value with the set difference threshold or difference range, and if the difference value exceeds the set threshold or range, it is determined to be a significant deviation.

[0015] Compared with the prior art, the present invention has the following beneficial effects: By constructing a smart home device state prediction model, the present invention can accurately predict the operating state of each smart device in the home within a period of time in the future, so as to formulate a collaborative operation strategy in advance, avoid disordered work and energy waste between devices, and significantly improve the overall operating efficiency of the smart home system. Through the device collaborative control decision model, the present invention can automatically generate preliminary collaborative operation instructions for each device according to the predicted future operating state of each smart device, thereby realizing intelligent collaborative control between devices. This control method not only responds quickly, but also can be adaptively adjusted according to user needs and changes in the home environment, providing a more personalized service experience. The introduction of the difference fusion model enables the present invention to monitor the actual operating state of each smart device in real time and compare it with the predicted results. Once a significant deviation is found, the difference fusion model will immediately recalculate the optimal collaborative operation instructions, optimize and adjust the preliminary instructions, thereby ensuring the accuracy and effectiveness of the collaborative operation instructions.

[0016] The present invention does not rely on a specific smart device brand or communication protocol, but achieves collaborative operation between devices through data analysis and machine learning technology. This design method enhances the compatibility and scalability of the smart home system, allowing users to more freely choose and combine smart devices of different brands and types to build a smart home system that meets their needs. Through the collaborative operation method of the present invention, the smart home system can more accurately meet the actual needs of users and changes in the home environment, and provide a more convenient, comfortable and safe living environment. This intelligent service experience not only improves user satisfaction, but also helps to promote the popularization and development of smart homes. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A working principle diagram of a collaborative operation method of a smart home system according to the present invention; Figure 2 A flowchart for building a smart home device status prediction model; Figure 3 Flowchart for building the difference fusion model. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1-3 The present invention provides a technical solution: a method for coordinated operation of a smart home system, the method comprising: Step A: Construct a smart home device state prediction model and a device collaborative control decision model.

[0020] By collecting user habit data and home environment data, the machine learning algorithm is used to train the smart home device status prediction model. At the same time, based on the prediction results, the device collaborative control decision model is designed to generate preliminary collaborative operation instructions. This step can accurately predict the future operating status of each smart device and provide a basis for device collaboration. At the same time, the preliminary collaborative operation instructions can ensure basic collaboration between devices and improve user experience.

[0021] The smart home device state prediction model is used to predict the operating state of each smart device in the home in the future. The model input includes user habit data and home environment data, which are processed by a deep learning algorithm to output the future operating state of each smart device. By comprehensively considering user habits and home environment, the model can more accurately predict the device state and provide a reliable basis for subsequent collaborative control.

[0022] The device collaborative control decision model is used to control the collaborative work of each intelligent device. Based on the predicted device status, preliminary collaborative operation instructions are generated using rules or reinforcement learning algorithms. The model can quickly generate collaborative instructions based on the prediction results, ensure the initial collaboration between devices, and improve the system response speed.

[0023] Step B: Build a difference fusion model to optimize the initial collaborative operation instructions.

[0024] By monitoring the actual operation status of the equipment in real time and comparing it with the predicted results, the difference value is calculated. These differences are input into the difference fusion model to optimize the preliminary collaborative operation instructions and generate the optimal collaborative operation instructions. The difference fusion model can correct the prediction error in real time and ensure the accuracy of the collaborative operation instructions, which helps to improve the overall operation efficiency of the smart home system and better meet user needs.

[0025] Step C: Continuously monitor the actual operating status of each smart home device and compare it with the predicted results.

[0026] Based on the smart home device state prediction model and device collaborative control decision model, combined with actual user habit data and actual home environment data, the preliminary collaborative operation instructions of each smart device are generated to control the collaborative operation of each smart device. The actual operation status of each smart device is continuously monitored and compared with the prediction results of the smart home device state prediction model. If significant deviations are found, the difference fusion model is triggered to recalculate the optimal collaborative operation instructions. Continuous monitoring and comparison can ensure that the system can detect and correct prediction errors in a timely manner and maintain collaborative work between devices. This helps to enhance the stability and reliability of the smart home system and provide users with a continuous high-quality service experience.

[0027] The present invention will be further described below in conjunction with Examples 1 to 3: Embodiment 1:

[0028] When constructing a smart home device status prediction model, the present invention uses a convolutional neural network (CNN) algorithm. The following are the specific training steps of the model: Step 1: Collect training data: Collect user habit data, which includes user behavior patterns, preference settings, and historical operation records in smart homes. These data reflect user usage habits and preferences.

[0029] Collect home environment data, including temperature, humidity, light intensity, occupancy distribution, and time information within the smart home, which provides environmental context for the model.

[0030] Collect the actual operating status data of each corresponding smart device, including the device's switch status, working mode, working parameters and actual working time. These data serve as the target of model prediction.

[0031] Step 2: Preprocess the training data: Data cleaning: For missing values, linear interpolation algorithm is used to fill them in to ensure the continuity of data.

[0032] For outliers, the threshold judgment method is used to regard data points that exceed the preset threshold range as outliers and replace them with the mean of their neighboring points to eliminate the impact of abnormal data on model training.

[0033] Data standardization: The Z-score standardization method is used to standardize the cleaned data. Specifically, each numerical data is subtracted from its mean and divided by its standard deviation, so that the processed data conforms to the standard normal distribution. The formula for standardization is: ,in, is the original data, is the mean of the original data, is the standard deviation of the original data, is the standardized data.

[0034] Step 3: Build the model and set hyperparameters: The initial structure of the smart home device status prediction model is constructed using the convolutional neural network (CNN) algorithm.

[0035] Setting the model's hyperparameters, including the number of convolutional layers, the size of the convolution kernel, the choice of activation function, the type of optimizer, and the learning rate, has an important impact on the performance and training speed of the model.

[0036] Step 4: Model training and optimization: The initial model is trained using the preprocessed data set. During the training process, the actual operating status data of each smart device is used as a label, and the model parameters are adjusted through an iterative optimization algorithm so that the model can accurately predict the operating status of each smart device in the home in the future.

[0037] During the training process, the cross-validation method is used to evaluate the model. The data set is divided into multiple subsets, and training and validation are performed on each subset to prevent the model from overfitting. The model with the best performance is selected as the final smart home device status prediction model.

[0038] Embodiment 2:

[0039] When constructing the equipment collaborative control decision model, the present invention adopts the deep deterministic policy gradient (DDPG) algorithm. The following are the specific training steps and implementation methods of the model: Step 1: Build the network structure of the DDPG algorithm: Build an actor network, which is responsible for generating preliminary collaborative operation instructions for each smart device based on the current state. These instructions are regarded as actions in the DDPG algorithm.

[0040] Construct a critic network, which is responsible for evaluating the value of the actions generated by the actor network, that is, evaluating whether the action can enable the smart home system to achieve a better collaborative operation state.

[0041] Step 2: Collect training data: As the state input, the predicted future operating state of each smart device output by the smart home device state prediction model is collected. These data include the device's switch state, working mode, working parameters and expected working time, etc.

[0042] As target actions (labels), we collect optimal collaborative operation instructions annotated by humans. These instructions are determined based on expert knowledge or user feedback, and indicate how each smart device should work together to achieve the best effect in a given state.

[0043] Step 3: Initialize network parameters and set hyperparameters: Initialize the parameters of the actor network and the critic network, including the weights and biases of the network layers.

[0044] Set the hyperparameters of the DDPG algorithm, such as the learning rate, discount factor, exploration noise size, experience replay buffer size, and target network update frequency.

[0045] Step 4: Model training and optimization: During the training process, the actor network generates actions (i.e., preliminary collaborative operation instructions) based on the current state (i.e., the predicted future operating state of each smart device).

[0046] The critic network evaluates the value of the action, that is, calculates the expected return under the action. Then, the parameters of the actor network and the critic network are updated by gradient descent, so that the actor network can generate actions that are closer to the optimal collaborative operation instructions.

[0047] Specifically, the actor network updates its parameters via a policy gradient method to maximize the expected reward; the critic network updates its parameters via a temporal difference learning method to accurately evaluate the value of an action.

[0048] Step 5: Repeat training and model evaluation: Repeat step 4 until the model's performance indicators (such as average reward, success rate, or user satisfaction) reach the preset standard.

[0049] During the training process, the performance of the model can be evaluated regularly, and the hyperparameters or network structure can be adjusted based on the evaluation results to further optimize the model.

[0050] Assume that the smart home system includes three devices: smart lighting, smart temperature control, and smart security. When training the device collaborative control decision model, the state input may include the predicted operating status of these devices (such as the brightness of the lighting, the temperature setting of the temperature control, and the alarm status of the security, etc.), as well as home environment data (such as time, personnel distribution, etc.). The target action is the optimal collaborative operation instruction manually labeled, such as "When there are people in the living room and the light is dim, turn on the lighting and adjust it to a moderate brightness; when the indoor temperature is higher than the set value, turn on the air conditioner to cool down; when the security system detects an abnormality, trigger an alarm and notify the user." Through the training of the DDPG algorithm, the device collaborative control decision model can learn to generate near-optimal collaborative operation instructions based on the current state, thereby realizing efficient collaborative operation of the smart home system.

[0051] Embodiment 3:

[0052] When constructing the difference fusion model, the present invention adopts the Random Forest (RF) algorithm. The following are the specific training steps of the model.

[0053] Step 1: Collect training data: Collect the differences between the predicted operating status output by the smart home device status prediction model and the actual operating status of the device obtained by real-time monitoring of the smart home system. These differences include device status differences (such as whether the switch status is consistent), working time differences (such as the deviation between the predicted working time and the actual working time), and working mode differences (such as the difference between the predicted working mode and the actual working mode).

[0054] At the same time, we collect manually annotated optimal collaborative operation instructions given the difference between the predicted operating state and the actual operating state. These instructions are determined based on expert knowledge or user feedback, indicating how each smart device should work together to achieve the best effect under specific differences, and serve as labels for training data.

[0055] Step 2: Build a random forest RF algorithm model: Build a random forest model, which consists of multiple decision trees, each of which independently makes classification or regression predictions for the input data.

[0056] Set the parameters of the random forest, including the number of trees (determines the complexity and stability of the model), the maximum depth (controls the growth depth of the tree to prevent overfitting), and the minimum number of sample splits (determines the minimum number of samples required for node splitting).

[0057] Step 3: Model training: The collected training data is used to train the random forest RF algorithm model. During the training process, the model continuously adjusts the parameters and structure of each decision tree through iterative learning to accurately predict the optimal collaborative operation instructions under given input differences.

[0058] During training, methods such as cross-validation are used to evaluate the performance of the model and adjust parameters to optimize the model.

[0059] Step 4: Training completion and model acquisition: After sufficient training, a difference fusion model is obtained, which can output the optimal collaborative operation instructions based on the difference between the input predicted operation state and the actual operation state.

[0060] Step 5: Determine whether the deviation is significant: Set a difference threshold or difference range, which is determined based on actual application scenarios and expert experience. During the model application process, calculate the difference between the actual operating state of the smart home device and the predicted operating state, and compare the difference with the set difference threshold or range. If the difference exceeds the set threshold or range, it is judged as a significant deviation, and a specific processing mechanism may need to be triggered, such as sending an alarm, adjusting the device operating state, or requesting manual intervention.

[0061] Take the smart lighting and smart temperature control devices in smart homes as an example. Assume that at a certain moment, the predicted operating state of the smart lighting device is "on, brightness 80%", while the actual operating state is "off"; at the same time, the predicted operating state of the smart temperature control device is "temperature setting 22°C", while the actual operating state is "temperature setting 25°C".

[0062] First, the difference between the predicted operating status and the actual operating status of these devices is calculated. For example, the state difference of smart lighting is "on-off", and the working difference is significant; the working mode difference of smart temperature control is "3°C temperature deviation".

[0063] These differences are then used as input features into the trained difference fusion model.

[0064] Based on these differences, the model outputs the optimal collaborative operation instructions, such as "Since the status of lighting equipment is significantly different, turn on the lighting immediately and adjust it to the appropriate brightness; since the temperature control equipment has a deviation of 3°C, gradually adjust the temperature to 22°C."

[0065] At the same time, it determines whether these differences exceed the set threshold or range. If so, the corresponding processing mechanism is triggered, such as sending a notification to the user or adjusting the operation strategy of the device.

[0066] The difference fusion model based on the random forest (RF) algorithm can accurately predict the optimal collaborative operation instructions according to the difference between the predicted operating status and the actual operating status of smart home devices, and determine whether the deviation is significant, thereby improving the collaborative operation efficiency and user satisfaction of the smart home system.

[0067] In the present invention, the following detailed description is made on the deviation judgment between the actual operating state and the predicted operating state of the smart home device, and the relationship between the deviation and the device state difference, working time difference and working mode difference: Deviation refers to the degree of inconsistency between the actual operating status of smart home devices and the predicted operating status. It is obtained by comparing the actual monitored operating status of the device with the operating status output by the prediction model. Deviation is reflected in many aspects, including differences in device status, working time, and working mode.

[0068] Device state discrepancy: refers to the inconsistency between the actual device state (such as switch state, working mode, etc.) and the predicted state. For example, the prediction model may predict that a smart light bulb should be turned on at a certain time in the future, but the actual detection shows that the light bulb is turned off, which constitutes a device state discrepancy.

[0069] Working time difference: refers to the deviation between the actual working time of the equipment and the predicted working time. For example, the prediction model may predict that the smart air conditioner will continue to operate in a certain period of time in the future, but the actual monitoring shows that the air conditioner does not continue to operate in this period of time, or the operating time is offset, which constitutes a working time difference.

[0070] Operation mode discrepancy: refers to the inconsistency between the actual operation mode of the device and the predicted operation mode. For example, the prediction model may predict that the smart washing machine will operate in a certain mode (such as the quick wash mode), but the washing machine is actually monitored to operate in another mode (such as the standard wash mode), which constitutes an operation mode discrepancy.

[0071] In order to quantify this deviation and determine whether it is significant, the present invention adopts a method of setting a difference threshold. Specifically, corresponding difference thresholds or difference ranges can be set for device status differences, working time differences, and working mode differences. When the calculated difference value (such as the number of inconsistent states, the duration of time deviation, or the degree of mode difference) exceeds the set threshold or range, it is determined to be a significant deviation.

[0072] For example, for differences in device status, a tolerance number of status inconsistencies can be set as a threshold; for differences in working hours, a tolerance range of time deviation can be set; for differences in working modes, a corresponding threshold can be set according to the degree of difference in the modes. When the actual monitored differences exceed these set thresholds or ranges, the difference fusion model is triggered to recalculate the optimal collaborative operation instructions to adjust the operating status of the smart home system, ensuring that the system can more accurately adapt to the actual needs of users and changes in the home environment.

[0073] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0074] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A collaborative operation method of a smart home system, characterized in that: The method comprises: Step A: Construct a smart home device state prediction model and a device collaborative control decision model; The smart home device state prediction model is used to predict the operating state of each smart device in the home within a period of time in the future, and its input data includes: User habit data, including user behavior patterns, preference settings, and historical operation records in smart homes; Home environment data, including temperature, humidity, light intensity, occupancy distribution and time information within the smart home; The output data are: the future operating status of each smart device, including the predicted switch status, predicted working mode, predicted working parameters and predicted working time of the device; The device collaborative control decision model is used to control the collaborative work of each smart device; its input data is: the predicted future operating state of each smart device output by the smart home device state prediction model; its output data is: the preliminary collaborative operation instructions of each smart device; Step B: constructing a difference fusion model, which is used to optimize the preliminary collaborative operation instructions; its input data is: the difference between the predicted operation state output by the smart home device state prediction model and the actual operation state of the device obtained by real-time monitoring of the smart home system, including device state difference, working time difference and working mode difference; its output data is: the optimal collaborative operation instruction of the smart home system after optimizing the preliminary collaborative operation instruction; Step C: Based on the smart home device state prediction model and the device collaborative control decision model, combined with actual user habit data and actual home environment data, generate preliminary collaborative operation instructions for each smart device to control the collaborative operation of each smart device, continuously monitor the actual operating status of each smart device, and compare it with the prediction results of the smart home device state prediction model. If significant deviations are found, the difference fusion model is immediately triggered to recalculate the optimal collaborative operation instructions.

2. The method for cooperative operation of a smart home system according to claim 1, characterized in that: In step A, a convolutional neural network (CNN) algorithm is used to build a smart home device status prediction model.

3. The method for cooperative operation of a smart home system according to claim 2, characterized in that: The training steps of the smart home device state prediction model include: Collect training data, including user habit data, home environment data, and corresponding actual operating status data of each smart device; Preprocess the collected training data, including data cleaning and data standardization; Use the convolutional neural network (CNN) algorithm to build the initial structure of the smart home device status prediction model and set the model's hyperparameters; The initial model is trained using the preprocessed data set, and the model parameters are adjusted through an iterative optimization algorithm so that the model can accurately predict the operating status of each smart device in the home in the future. The labels of the training data are the actual operating status data of each smart device. During the training process, the cross-validation method was used to evaluate the model to prevent overfitting, and the model with the best performance was selected as the final smart home device status prediction model.

4. The method for cooperative operation of a smart home system according to claim 3, characterized in that: The data cleaning method comprises: For missing values, linear interpolation algorithm is used to fill them; For outliers, the threshold judgment method is used to regard data points that exceed the preset threshold range as outliers and replace them with the mean of their neighboring points.

5. The method for cooperative operation of a smart home system according to claim 3, characterized in that: The method for data standardization includes: using the Z-score standardization method to standardize the cleaned data, specifically: subtracting the mean of each numerical data and dividing it by its standard deviation, so that the processed data conforms to the standard normal distribution. The formula for standardization is: ,in, is the original data, is the mean of the original data, is the standard deviation of the original data, is the standardized data.

6. The method for cooperative operation of a smart home system according to claim 3, characterized in that: In step A, a deep deterministic policy gradient (DDPG) algorithm is used to construct a device collaborative control decision model.

7. A method for cooperative operation of a smart home system according to claim 6, characterized in that: The training steps of the equipment collaborative control decision model include: Step 1: Construct the network structure of the deep deterministic policy gradient DDPG algorithm, including the actor network and the critic network; the actor network is responsible for generating preliminary collaborative operation instructions for each intelligent device to represent the action; the critic network is responsible for evaluating the value of the action generated by the actor network; Step 2: Collect training data, including the predicted future operating state of each smart device output by the smart home device state prediction model as state input, and the corresponding optimal collaborative operation instructions annotated by humans as labels to represent the target action; Step 3: Initialize the parameters of the actor network and the critic network, and set the hyperparameters of the DDPG algorithm; Step 4: During the training process, the actor network generates an action based on the current state, the critic network evaluates the value of the action, and updates the parameters of the actor network and the critic network through the gradient descent method, so that the actor network generates actions that are closer to the optimal collaborative operation instructions; Step 5: Repeat step 4 until the performance index of the model reaches the preset standard, and a trained equipment collaborative control decision model is obtained.

8. The method for cooperative operation of a smart home system according to claim 7, characterized in that: In the step B, a random forest RF algorithm is used to construct a difference fusion model.

9. The method for cooperative operation of a smart home system according to claim 8, characterized in that: The training steps of the difference fusion model include: Collect training data, including the difference between the predicted operating state output by the smart home device state prediction model and the actual operating state of the device obtained by real-time monitoring of the smart home system, including differences in device state, working time, and working mode; at the same time, collect manually annotated optimal collaborative operation instructions under the given difference between the predicted operating state and the actual operating state as labels for the training data; Build a random forest RF algorithm model and set the parameters of the random forest, including the number of trees, maximum depth, and minimum number of sample splits; Use the training data to train the random forest RF algorithm model, and through iterative learning, enable the model to accurately predict the optimal collaborative operation instructions under given input differences; After the training is completed, the difference fusion model is obtained.

10. The method for cooperative operation of a smart home system according to claim 1, characterized in that: The method for determining whether the deviation between the actual operating state and the predicted operating state of the smart home device is significant is: setting a difference threshold, comparing the calculated difference value with the set difference threshold or difference range, and if the difference value exceeds the set threshold or range, it is determined to be a significant deviation.

Images