A control system and method for Internet of Things equipment based on big data

Abstract

The present invention discloses an Internet of Things device control system and method based on big data, comprising a distributed data acquisition module, a data normalization module, a data storage and management module, a big data mining module, a control and decision module, a secure communication module, and an Internet of Things device control module. The distributed data acquisition module is used for real-time data acquisition, the data normalization module is used for data standardization processing, the data storage and management module is used for data management and storage, the big data mining module is used for information mining of big data, the control and decision module is used for control strategy formulation and decision execution, the secure communication module is used for data secure transmission, and the Internet of Things device control module is used for controlling Internet of Things devices. The present invention discloses an Internet of Things device control system and method based on big data, proposes a big data mining algorithm based on fuzzy clustering to mine information on big data, and proposes a big data analysis control algorithm based on reinforcement learning to control Internet of Things devices.

Description

Technical Field

[0001] The present invention relates to the fields of big data mining, reinforcement learning and big data analysis, and specifically to a control system and method for Internet of Things devices based on big data. Background Art

[0002] Big data mining technology is a technology that analyzes massive amounts of data to discover hidden patterns, associations, and regularities. It aims to discover hidden patterns, associations, and trends from large-scale data sets to provide valuable information and decision support. By applying statistics, machine learning, and data visualization technology, big data mining can help users discover regularities in data and obtain useful information from it, thereby achieving intelligent management and control of IoT devices. In IoT device control systems, big data mining technology can help the system extract valuable information from massive amounts of device data for optimizing device control strategies and predicting device status.

[0003] Big data analysis technology is a data processing and analysis method that aims to analyze effective information extracted from large-scale data sets to support decision-making and problem solving. In the control system of IoT devices, the optimal control strategy is obtained by combining reinforcement learning technology with big data analysis technology, aiming to solve intelligent decision-making problems in complex environments. Reinforcement learning technology can optimize the control strategy of devices through interactive learning with IoT devices, and realize intelligent management and control of devices. Big data analysis technology can analyze large amounts of data from different devices, discover correlations and trends between data, and provide support and guidance for intelligent control of equipment.

[0004] However, the existing IoT device control systems and methods based on big data lack the support of big data mining and analysis technology. The system cannot effectively process massive data, resulting in the inability to fully mine the potential information and value in the data. In addition, without the application of reinforcement learning intelligent algorithms, the system cannot dynamically adjust the control strategy according to environmental changes and the status of IoT devices, resulting in poor control effects. Summary of the invention

[0005] The purpose of the present invention is to provide an Internet of Things device control system and method based on big data, so as to solve the problem that the existing Internet of Things device control system and method based on big data proposed in the above background technology lack big data mining and analysis technical support, the system cannot effectively process massive data, resulting in the inability to fully mine the potential information and value in the data, and the system cannot dynamically adjust the control strategy according to environmental changes and the status of the Internet of Things device without applying reinforcement learning intelligent algorithms, resulting in poor control effect.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a control system and method for Internet of Things devices based on big data, including a distributed data acquisition module, a data normalization module, a data storage and management module, a big data mining module, a control and decision module, a security communication module, and an Internet of Things device control module, characterized in that: the distributed data acquisition module is used to collect data from various Internet of Things devices, including sensor data and device status, and transmit the data to the control system, the data normalization module is used to standardize the format and uniformly process the collected data to ensure the consistency and comparability of the data, the data storage and management module is used to store the data in a suitable storage medium, and organize, index, access control, backup and restore the data to ensure the consistency, integrity and availability of the data, and the big data mining module proposes big data mining based on fuzzy clustering The algorithm is used to analyze and mine the effective information of the stored data to discover the patterns, trends and correlations hidden behind the big data and provide support for subsequent decision-making. The control and decision-making module includes a big data analysis control unit and a big data decision execution unit. The big data analysis control unit proposes a big data analysis control algorithm based on reinforcement learning to conduct in-depth analysis of the effective information of the big data extracted by the big data mining module, identify the status, abnormal conditions and trend changes of the Internet of Things devices, and provide real-time control suggestions. The big data decision execution unit is used to receive the results from the analysis control unit and formulate corresponding control strategies and decision-making plans. The secure communication module is used to ensure the security of internal and external communications of the system, including data encryption, identity authentication, and access control. The Internet of Things device control module is used to communicate and control with Internet of Things devices, including sending direct control instructions and receiving device status feedback.

[0007] Preferably, the distributed data acquisition module works simultaneously through multiple data acquisition nodes to efficiently collect data from various IoT devices and transmits the data to the data storage and management module in real time, thereby ensuring the efficiency and real-time nature of data acquisition and avoiding single point failures affecting the overall system operation.

[0008] Preferably, the data normalization module ensures the consistency and comparability of data during storage and processing by normalizing data formats and standardizing data fields, reduces the complexity and error rate of data processing, and ensures that data of different types and formats can be processed in a unified and standardized manner for subsequent data storage, management, and analysis.

[0009] Preferably, the data storage and management module securely stores the collected data, and effectively manages the big data through data indexing, permission control, and data backup functions, thereby ensuring the effectiveness and reliability of data organization, management, and maintenance.

[0010] Preferably, the big data mining module proposes a big data mining algorithm based on fuzzy clustering, deeply mines the effective information of the stored big data, and reveals the potential information and regularity in the data through pattern recognition, association analysis and anomaly detection, thereby providing support and insights for decision-making.

[0011] Specifically, first, the original large data set X processed by the data normalization module is divided into data blocks of equal size according to the IoT devices. Each data block is an independent unit and can be processed independently, which helps to improve the efficiency and speed of the algorithm. The formula for data block division is as follows:

[0012] X={X1,X2,...,X M}={X i |i=1,...,M}

[0013] Among them, X represents the original large data set, X1, X2, ..., X M Represented as a data block, X i It is represented as the i-th data block after division, i is represented as the index of the number of data blocks, and M is represented as the total number of data blocks. By dividing the original large data set into M data blocks of equal size, the algorithm can efficiently process large-scale data in the subsequent processing process and reduce the computational complexity; secondly, incremental fuzzy clustering is performed on each data block to discover the patterns and association rules therein, so as to mine valuable information in big data. By optimizing the calculation of the membership between the data point and the cluster center to the calculation based on the distance between the data point and the cluster center, it can well reflect the similarity between the data point and the cluster center and improve the accuracy of clustering. The specific formula is as follows:

[0014]

[0015] Among them, q sv Represented as data point x s With cluster center c v The degree of membership between xy The larger the value, the stronger the data point x s More likely to belong to cluster center c v The cluster where d sv Represented as data point x s With cluster center c v The distance between ls Represented as data point x s With cluster center c l The distance between them, η is represented by the fuzzy coefficient, which is used to control the fuzziness of the membership, s is represented by the index of the data point, v is represented by the index of the cluster center, l∈v is represented by the index of the lth cluster center, and the data point xs represents the data in each data block, min(·) represents the minimum value function, and the specific formula is:

[0016] X={x,|s=1,...,S}

[0017] Among them, S represents the total number of data points, and the cluster center c is calculated by weighted average method. v Initialization is performed so that the algorithm can take into account the importance of data points and improve the accuracy of cluster centers. The specific formula is as follows:

[0018]

[0019] Among them, λ s Represented as data point x s The weight is used to consider the importance of the data point. τ is expressed as a constant coefficient to avoid the situation where the denominator is zero and ensure the stability of the calculation. By calculating the weight and membership of the data point, and then performing weighted summation on each data point, the updated cluster center is finally obtained; then, according to the data block X i The distance between the cluster center c at the previous moment t-1 v (t-1) The most recent data point And the data points Redistribute to the corresponding cluster center to update the cluster center c at time t v (t) position, while maintaining the fuzzy clustering characteristics, better use of the clustering results of the previous moment to avoid the emergence of local optimal solutions. The specific formula for updating the cluster center position is as follows:

[0020]

[0021] Among them, c v (t) represents the vth cluster center after updating at time t, Represented as data block X i The distance between the cluster center c at the previous moment t-1 v (t-1) the most recent data point, Represented as data points With the initial cluster center c v The distance between them is dynamically updated to reflect the contribution of the data point to the cluster center. The cluster center weight update formula is specifically expressed as:

[0022]

[0023] Among them, σ represents the coefficient that controls the smoothness of weight update. σ makes the weight update smoother and reduces λ sThe drastic change of weights during the update process is conducive to the stability and convergence speed of the algorithm; finally, the data blocks corresponding to each IoT device are integrated through multiple intermediate nodes to solve the problem that a single transfer point cannot completely contain all information, thereby improving the reliability and accuracy of information mining, clustering each data block and transfer point set, and then re-clustering to obtain the updated transfer point. The calculation formula of the transfer point coefficient of each IoT device is as follows:

[0024]

[0025] in, It is represented as the transfer point coefficient between the ith IoT device and the jth IoT device at time t, j∈i is the index of the jth IoT device, Represented as data block X at time t i and X j The inverse matrix of the kernel function values ​​between Represented as data block X at time t i and X j The kernel function value between It is expressed as the transfer point coefficient between the i-th IoT device and the j-th IoT device at time t-1. By mapping the new cluster center to the data block space, the optimal transfer point can be selected. The specific calculation formula is expressed as:

[0026]

[0027] Among them, Φ(c v (t),i) represents the new cluster center of the jth class at time t, Φ(X i ) represents the data block X i All data points in the data are mapped to a function in a high-dimensional space. The design of the entire algorithm fully utilizes the information and value hidden in big data to achieve intelligent control and optimization of IoT devices. Through intelligent data mining, the algorithm can discover potential patterns and laws in the data and provide a more accurate reference for decision-making. The multi-transmission point selection method further improves the accuracy and efficiency of data analysis, ensuring the full utilization and value mining of big data.

[0028] Preferably, the control and decision-making module includes a big data analysis control unit, which proposes a big data analysis control algorithm based on reinforcement learning, performs real-time analysis on the effective information of the big data extracted by the big data mining module, identifies patterns, trends and abnormal situations in the data, and updates control recommendations in a timely manner to support the system's intelligent control and decision-making of IoT devices.

[0029] Specifically, the big data analysis control algorithm based on reinforcement learning is as follows: First, determine the state space and executable action space of the IoT device. The state space describes all the states that the IoT device may be in, while the action space describes the operations that can be performed on the IoT device. The specific expression formulas of the state space and the action space are:

[0030]

[0031] Among them, S represents the state space of IoT devices, Represents the IoT device status, represents the mth state of the ith IoT device, i represents the number index of IoT devices, I represents the total number of IoT devices, m represents the state index of each IoT device, A represents the action space of IoT devices, Represents the actions of IoT devices, It is represented as the mth action of the ith IoT device, and n is the action index of each IoT device. By defining the state space and action space, the control problem of the IoT device is transformed into a reinforcement learning problem, which provides a basis for the subsequent establishment of strategies. Then, the reward function is used to evaluate the execution quality of each action to guide the learning process of the reinforcement learning algorithm. The reward function can provide feedback signals under the reinforcement learning framework, which helps the control system understand the quality of the behavior, so that the algorithm can learn to take the best action in a given environment. The reward function is constructed as a function of the state of the system at each time step and the action performed. The specific formula is expressed as:

[0032]

[0033] in, It is expressed as the state at time t Next action The reward function is It is expressed as the state at time t Next action The energy efficiency reward value obtained after , w1 represents the weight coefficient of the energy efficiency reward part, It is expressed as the state at time t Next action The performance optimization reward value obtained after , w2 is expressed as the weight coefficient of the performance optimization reward part, It is expressed as the state at time t Next action The communication efficiency reward value obtained after that, w3 is expressed as the weight coefficient of the communication efficiency reward part, through the reward function Comprehensively consider energy consumption, performance and communication efficiency factors to guide IoT devices to choose appropriate actions in a distributed environment to optimize the overall control strategy; then, the Actor-Critic structure is used to build a reinforcement learning model, where the Actor model is used to learn the strategy of selecting actions in a given state, and the Critic model is used to learn the expected return of performing an action in a given state. The specific formulas of the Actor model and the Critic model are expressed as follows:

[0034]

[0035] Among them, O Actor (θ) represents the optimization objective function of the Actor model, θ represents the parameter of the Actor model, is expressed as the expected function, min(·) is expressed as the minimum function, It is expressed as a given state at time t under the Actor model parameter θ Select action The probability of It is expressed as a given state at time t-1 under the Actor model parameter θ Select action The probability of Expressed as the advantage function at time t, it is used to measure the execution of actions Compared with the average level, clip is represented by the clipping function, which is used to limit the range of the ratio, ∈ is represented by the clipping parameter, which is used to control the range of clipping, O Critic (μ) is the optimization objective function of the Critic model, which is used to update the parameters of the Critic model. Expressed as the Critic model for the state at time t The action value estimation function is represented by |·|, which is an absolute value operation, and μ is a parameter of the Critic model. Finally, by optimizing the objective function, the parameters of the Actor model and the Critic model are updated to improve the strategy and control the IoT device. The objective function is optimized using the gradient descent method to update the parameters of the two models. The specific calculation formula is expressed as:

[0036]

[0037] Among them, η θ Expressed as the learning rate of the Actor model, η μ Expressed as the learning rate of the Critic model, It is expressed as a gradient calculation function. By adopting the Actor-Critic structure, it can effectively learn the strategy of selecting the optimal action under a given state, and can provide feedback based on the value of the action to guide the optimization of the strategy. By combining the reinforcement learning architecture with multiple IoT devices to share learning experience and model parameters, the system can learn and control multiple devices more robustly. This cooperative learning method can not only improve the robustness of the system, but also achieve adaptive control of different devices.

[0038] Preferably, the control and decision-making module includes a big data decision execution unit, which formulates a specific control strategy based on the big data analysis results and sends control instructions to the security communication module to achieve remote control and management of the Internet of Things devices.

[0039] Preferably, the secure communication module encrypts and authenticates data through a secure communication protocol to prevent unauthorized access, theft and tampering of the data, protect the confidentiality and integrity of the communication data, and ensure the security and reliability of control instructions and data transmission.

[0040] Preferably, the IoT device control module communicates with the IoT device, sends control instructions to the device according to the control decision made by the control and decision module, and remotely controls and manages the IoT device to improve the management efficiency and response speed of the IoT device.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The big data mining module proposes a big data mining algorithm based on fuzzy clustering. The algorithm performs intelligent data mining through fuzzy clustering and uses a multi-transfer point selection method to extract valuable information from huge data sets to provide support for the intelligent control of IoT devices. First, the algorithm uses fuzzy clustering technology to intelligently divide and classify big data. In the process of fuzzy clustering, each data point is assigned a set of membership, which reflects the degree of belonging of the data point to each cluster center. Through iterative update optimization, the algorithm continuously updates the cluster center and membership, thereby realizing the intelligent classification of big data. This process aims to discover the potential patterns and laws in the data and provide a basis for subsequent data analysis and application. Secondly, the multi-transfer point selection technology is used to map the cluster center to the data block space in order to select the best transfer point. The selection of multiple transfer points is based on The distance and similarity between data points and cluster centers are used to ensure that the transfer points can effectively represent each cluster center. By mapping the cluster centers to the data block space, the algorithm can accurately capture the correlation and characteristics between the data, providing a more reliable basis for subsequent data analysis and decision-making. The design of the entire algorithm makes full use of the information and value hidden in big data to achieve intelligent control and optimization of IoT devices. Through intelligent data mining, the algorithm can discover potential patterns and laws in the data and provide a more accurate reference for decision-making. The multi-transfer point selection technology further improves the accuracy and efficiency of data analysis, ensuring the full utilization and value mining of big data. In general, the big data mining algorithm based on fuzzy clustering can not only realize the intelligent classification and categorization of big data, but also provide strong support for the intelligent control and decision-making of IoT devices.

[0043] 2. Big Data Analysis Control Unit A big data analysis control algorithm based on reinforcement learning is proposed. This algorithm combines reinforcement learning technology and big data analysis. By learning and optimizing control strategies, it realizes intelligent management of IoT devices. First, the algorithm clearly defines the state space and action space of IoT devices so that they can be modeled under the reinforcement learning framework. Such a definition enables the system to accurately understand the current state of IoT devices and the operations that can be performed, providing a basis for intelligent control. In the process of building a reinforcement learning model, the Actor-Critic structure is adopted, in which the Actor model is responsible for learning strategies and the Critic model is responsible for evaluating the value of actions. The Actor model learns the strategy of selecting the optimal action under a given state, while the Critic model guides the optimization process of the Actor model by evaluating the difference between the reward of an action and the value of an action. This structure can effectively learn the optimal control strategy and can optimize the control strategy according to the value of the action. The value is fed back to guide the improvement of the strategy. In terms of optimizing the objective function, the amplitude of the strategy update is controlled to prevent the strategy from changing too much, thereby improving the stability of learning. The parameters of the Actor model and the Critic model are updated through the stochastic gradient descent method, so that the output action probability of the Actor model can approach the optimal strategy, thereby maximizing the expected return. By jointly reinforcing the learning architecture of IoT devices, multiple IoT devices share learning experience and model parameters, so that the system can learn and control multiple devices more robustly. This cooperative learning method not only improves the robustness of the system, but also can achieve adaptive control of different devices. In summary, the big data analysis control algorithm based on reinforcement learning provides an innovative solution for the intelligent control of IoT devices. By learning and optimizing control strategies, intelligent management of multiple IoT devices is achieved, thereby improving the robustness and adaptability of the system, and providing strong support for the development and application of IoT technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The invention is further illustrated by the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0045] Figure 1 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0046] 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.

[0047] See also Figure 1 The present invention provides an Internet of Things device control system and method based on big data, including a distributed data acquisition module, a data normalization module, a data storage and management module, a big data mining module, a control and decision module, a security communication module, and an Internet of Things device control module, characterized in that: the distributed data acquisition module is used to collect data from various Internet of Things devices, including sensor data and device status, and transmit the data to the control system; the data normalization module is used to standardize and uniformly process the format of the collected data to ensure the consistency and comparability of the data; the data storage and management module is used to store the data in a suitable storage medium, and organize, index, access control, backup and restore the data to ensure the consistency, integrity and availability of the data; the big data mining module proposes a big data mining algorithm based on fuzzy clustering for storage The effective information of the data is analyzed and mined to discover the patterns, trends and correlations hidden behind the big data, and provide support for subsequent decision-making. The control and decision-making module includes a big data analysis control unit and a big data decision execution unit. The big data analysis control unit proposes a big data analysis control algorithm based on reinforcement learning to conduct in-depth analysis of the effective information of the big data extracted by the big data mining module, identify the status, abnormal conditions and trend changes of IoT devices, and provide real-time control suggestions. The big data decision execution unit is used to receive the results from the analysis control unit and formulate corresponding control strategies and decision-making plans. The secure communication module is used to ensure the security of internal and external communications of the system, including data encryption, identity authentication, and access control. The IoT device control module is used to communicate and control IoT devices, including sending direct control instructions and receiving device status feedback.

[0048] See also Figure 1 ,Furthermore, the distributed data acquisition module works simultaneously through multiple ,data acquisition nodes, efficiently collects data from various IoT devices, and transmits ,data storage and management module in real time, ensuring the ,efficiency and real time of data acquisition and avoiding ,single point failure affecting the overall system operation.

[0049] See also Figure 1,Furthermore, the data normalization module ensures the consistency and ,comparability of data during storage and processing by normalizing the data format and ,standardizing data fields, reducing the complexity and error rate of data ,processing, and ensuring that data of different types and formats can be ,processed uniformly and standardized for subsequent data storage, ,management and analysis.

[0050] See also Figure 1 ,Furthermore, the data storage and management module securely stores the collected ,data, and effectively manages the big data through data indexing, permission control, and ,data backup functions, ensuring the effectiveness and reliability of data organization, ,management and maintenance.

[0051] See also Figure 1 ,Furthermore, the big data mining module proposes a big data mining ,algorithm based on fuzzy clustering, which deeply mines the effective ,information of the stored big data through pattern recognition, association ,analysis and anomaly detection methods to reveal the potential ,information and regularity in the data, thus providing support and ,insight for decision-making.

[0052] See also Figure 1 ,Furthermore, the big data mining algorithm based on fuzzy clustering is as follows: ,First, the original big data set X processed by the data normalization module is ,divided into data blocks of equal size according to the IoT devices. ,Each data block is an independent unit and can be processed independently, ,which helps to improve the efficiency and speed of the algorithm. ,The formula for data block division is as follows:

[0053] X={X1,X2,...,X M}={X i |i=1,...,M}

[0054] Among them, X represents the original large data set, X1, X2, ..., X M Represented as a data block, X i It is represented as the i-th data block after division, i is represented as the index of the number of data blocks, and M is represented as the total number of data blocks. By dividing the original large data set into M data blocks of equal size, the algorithm can efficiently process large-scale data in the subsequent processing process and reduce the computational complexity; secondly, incremental fuzzy clustering is performed on each data block to discover the patterns and association rules therein, so as to mine valuable information in big data. By optimizing the calculation of the membership between the data point and the cluster center to the calculation based on the distance between the data point and the cluster center, it can well reflect the similarity between the data point and the cluster center and improve the accuracy of clustering. The specific formula is as follows:

[0055]

[0056] Among them, q svRepresented as data point x s With cluster center c v The degree of membership between xy The larger the value, the stronger the data point x s More likely to belong to cluster center c v The cluster where d sv Represented as data point x s With cluster center c v The distance between ls Represented as data point x s With cluster center c l The distance between them, η is represented by the fuzzy coefficient, which is used to control the fuzziness of the membership, s is represented by the index of the data point, v is represented by the index of the cluster center, l∈v is represented by the index of the lth cluster center, and the data point x s represents the data in each data block, min(·) represents the minimum value function, and the specific formula is:

[0057] X i ={x s |s=1,...,S}

[0058] Among them, S represents the total number of data points, and the cluster center c is calculated by weighted average method. v Initialization is performed so that the algorithm can take into account the importance of data points and improve the accuracy of cluster centers. The specific formula is as follows:

[0059]

[0060] Among them, λ s Represented as data point x s The weight is used to consider the importance of the data point. τ is expressed as a constant coefficient to avoid the situation where the denominator is zero and ensure the stability of the calculation. By calculating the weight and membership of the data point, and then performing weighted summation on each data point, the updated cluster center is finally obtained; then, according to the data block X i The distance between the cluster center c at the previous moment t-1 v (t-1) The most recent data point And the data points Redistribute to the corresponding cluster center to update the cluster center c at time t v (t) position, while maintaining the fuzzy clustering characteristics, better use of the clustering results of the previous moment to avoid the emergence of local optimal solutions. The specific formula for updating the cluster center position is as follows:

[0061]

[0062] Among them, c v(t) represents the vth cluster center after updating at time t, Represented as data block X i The distance between the cluster center c at the previous moment t-1 v (t-1) the most recent data point, Represented as data points With the initial cluster center c v The distance between them is dynamically updated to reflect the contribution of the data point to the cluster center. The cluster center weight update formula is specifically expressed as:

[0063]

[0064] Among them, σ represents the coefficient that controls the smoothness of weight update. σ makes the weight update smoother and reduces λ s The drastic change of weights during the update process is conducive to the stability and convergence speed of the algorithm; finally, the data blocks corresponding to each IoT device are integrated through multiple intermediate nodes to solve the problem that a single transfer point cannot completely contain all information, thereby improving the reliability and accuracy of information mining, clustering each data block and transfer point set, and then re-clustering to obtain the updated transfer point. The calculation formula of the transfer point coefficient of each IoT device is as follows:

[0065]

[0066] in, It is represented as the transfer point coefficient between the ith IoT device and the jth IoT device at time t, j∈i is the index of the jth IoT device, Represented as data block X at time t i and X j The inverse matrix of the kernel function values ​​between Represented as data block X at time t i and X j The kernel function value between It is expressed as the transfer point coefficient between the i-th IoT device and the j-th IoT device at time t-1. By mapping the new cluster center to the data block space, the optimal transfer point can be selected. The specific calculation formula is expressed as:

[0067]

[0068] Among them, Φ(c v (t),i) represents the new cluster center of the jth class at time t, Φ(X i ) represents the data block X iAll data points in the data are mapped to a function in a high-dimensional space. The design of the entire algorithm fully utilizes the information and value hidden in big data to achieve intelligent control and optimization of IoT devices. Through intelligent data mining, the algorithm can discover potential patterns and laws in the data and provide a more accurate reference for decision-making. The multi-transmission point selection method further improves the accuracy and efficiency of data analysis, ensuring the full utilization and value mining of big data.

[0069] See also Figure 1 ,Furthermore, the control and decision making module includes a big data analysis ,control unit, which proposes a big data analysis control algorithm based on reinforcement learning, through real-time analysis of big data extracted by the big data mining module, identifying ,patterns, trends and abnormal situations in the data, and timely updating the ,control suggestions to support the system in intelligent control and decision-making of ,IoT devices.

[0070] See also Figure 1 ,Furthermore, the big data analysis control algorithm based on reinforcement learning is as follows: First, determine the state space and executable action space of the IoT device. The state space describes all the states that the IoT device may be in, while the action space describes the operations that can be performed on the IoT device. The specific expression formulas of the state space and action space are:

[0071]

[0072] Among them, S represents the state space of IoT devices, Represents the IoT device status, represents the mth state of the ith IoT device, i represents the number index of IoT devices, I represents the total number of IoT devices, m represents the state index of each IoT device, A represents the action space of IoT devices, Represents the actions of IoT devices, It is represented as the mth action of the ith IoT device, and n is the action index of each IoT device. By defining the state space and action space, the control problem of the IoT device is transformed into a reinforcement learning problem, which provides a basis for the subsequent establishment of strategies. Then, the reward function is used to evaluate the execution quality of each action to guide the learning process of the reinforcement learning algorithm. The reward function can provide feedback signals under the reinforcement learning framework, which helps the control system understand the quality of the behavior, so that the algorithm can learn to take the best action in a given environment. The reward function is constructed as a function of the state of the system at each time step and the action performed. The specific formula is expressed as:

[0073]

[0074] in, It is expressed as the state at time t Next action The reward function is It is expressed as the state at time t Next action The energy efficiency reward value obtained after , w1 represents the weight coefficient of the energy efficiency reward part, It is expressed as the state at time t Next action The performance optimization reward value obtained after , w2 is expressed as the weight coefficient of the performance optimization reward part, It is expressed as the state at time t Next action The communication efficiency reward value obtained after that, w3 is expressed as the weight coefficient of the communication efficiency reward part, through the reward function Comprehensively consider energy consumption, performance and communication efficiency factors to guide IoT devices to choose appropriate actions in a distributed environment to optimize the overall control strategy; then, the Actor-Critic structure is used to build a reinforcement learning model, where the Actor model is used to learn the strategy of selecting actions in a given state, and the Critic model is used to learn the expected return of performing an action in a given state. The specific formulas of the Actor model and the Critic model are expressed as follows:

[0075]

[0076] Among them, O Actor (θ) represents the optimization objective function of the Actor model, θ represents the parameter of the Actor model, is expressed as the expected function, min(·) is expressed as the minimum function, It is expressed as a given state at time t under the Actor model parameter θ Select action The probability of It is expressed as a given state at time t-1 under the Actor model parameter θ Select action The probability of Expressed as the advantage function at time t, it is used to measure the execution of actions Compared with the average level, clip is represented by the clipping function, which is used to limit the range of the ratio, ∈ is represented by the clipping parameter, which is used to control the range of clipping, O Critic (μ) is the optimization objective function of the Critic model, which is used to update the parameters of the Critic model. Expressed as the Critic model for the state at time t The action value estimation function is represented by |·|, which is an absolute value operation, and μ is a parameter of the Critic model. Finally, by optimizing the objective function, the parameters of the Actor model and the Critic model are updated to improve the strategy and control the IoT device. The objective function is optimized using the gradient descent method to update the parameters of the two models. The specific calculation formula is expressed as:

[0077]

[0078] Among them, η θ Expressed as the learning rate of the Actor model, η μ Expressed as the learning rate of the Gritic model, It is expressed as a gradient calculation function. By adopting the Actor-Gritic structure, it can effectively learn the strategy of selecting the optimal action under a given state, and can provide feedback based on the value of the action to guide the optimization of the strategy. By combining the reinforcement learning architecture with multiple IoT devices to share learning experience and model parameters, the system can learn and control multiple devices more robustly. This cooperative learning method can not only improve the robustness of the system, but also achieve adaptive control of different devices. First, determine the state space and executable action space of the IoT device. The state space describes all the states that the IoT device may be in, while the action space describes the operations that can be performed on the IoT device. The specific expression formulas of the state space and action space are:

[0079]

[0080] Among them, S represents the state space of IoT devices, Represents the IoT device status, represents the mth state of the ith IoT device, i represents the number index of IoT devices, i represents the total number of IoT devices, m represents the state index of each IoT device, A is the action space of IoT devices, Represents the actions of IoT devices, It is represented as the mth action of the ith IoT device, and n is the action index of each IoT device. By defining the state space and action space, the control problem of the IoT device is transformed into a reinforcement learning problem, which provides a basis for the subsequent establishment of strategies. Then, the reward function is used to evaluate the execution quality of each action to guide the learning process of the reinforcement learning algorithm. The reward function can provide feedback signals under the reinforcement learning framework, which helps the control system understand the quality of the behavior, so that the algorithm can learn to take the best action in a given environment. The reward function is constructed as a function of the state of the system at each time step and the action performed. The specific formula is expressed as:

[0081]

[0082] in, It is expressed as the state at time t Next action The reward function is It is expressed as the state at time t Next action The energy efficiency reward value obtained after , w1 represents the weight coefficient of the energy efficiency reward part, It is expressed as the state at time t Next action The performance optimization reward value obtained after , w2 is expressed as the weight coefficient of the performance optimization reward part, It is expressed as the state at time t Next action The communication efficiency reward value obtained after that, w3 is expressed as the weight coefficient of the communication efficiency reward part, through the reward function Comprehensively consider energy consumption, performance and communication efficiency factors to guide IoT devices to choose appropriate actions in a distributed environment to optimize the overall control strategy; then, the Actor-Critic structure is used to build a reinforcement learning model, where the Actor model is used to learn the strategy of selecting actions in a given state, and the Critic model is used to learn the expected return of performing an action in a given state. The specific formulas of the Actor model and the Critic model are expressed as follows:

[0083]

[0084]

[0085] Among them, O Actor (θ) represents the optimization objective function of the Actor model, θ represents the parameter of the Actor model, is expressed as the expected function, min(·) is expressed as the minimum function, It is expressed as a given state at time t under the Actor model parameter θ Select action The probability of It is expressed as the state given at time t-1 under the Acotr model parameter θ Select action The probability of Expressed as the advantage function at time t, it is used to measure the execution of actions Compared with the average level, clip is represented by the clipping function, which is used to limit the range of the ratio, ∈ is represented by the clipping parameter, which is used to control the range of clipping, O Critic(μ) is the optimization objective function of the Critic model, which is used to update the parameters of the Critic model. Expressed as the Critic model for the state at time t The action value estimation function is represented by |·|, which is an absolute value operation, and μ is a parameter of the Critic model. Finally, by optimizing the objective function, the parameters of the Actor model and the Critic model are updated to improve the strategy and control the IoT device. The objective function is optimized using the gradient descent method to update the parameters of the two models. The specific calculation formula is expressed as:

[0086]

[0087] Among them, η θ Expressed as the learning rate of the Actor model, η μ Expressed as the learning rate of the Critic model, It is expressed as a gradient calculation function. By adopting the Actor-Crititc structure, it can effectively learn the strategy of selecting the optimal action under a given state, and can provide feedback based on the value of the action to guide the optimization of the strategy. By combining the reinforcement learning architecture with multiple IoT devices to share learning experience and model parameters, the system can learn and control multiple devices more robustly. This cooperative learning method can not only improve the robustness of the system, but also achieve adaptive control of different devices.

[0088] See also Figure 1 ,Furthermore, the control and decision module includes a big data decision ,execution unit, which formulates specific control strategies according to the big data analysis ,results, and sends control instructions to the secure communication ,module to achieve remote control and management of the IoT devices.

[0089] See also Figure 1 ,Furthermore, the secure communication module encrypts and authenticates the data through the ,secure communication protocol to prevent the data from being accessed, stolen and tampered with without ,authorization, protecting the confidentiality and integrity of the ,communication data, and ensuring the security and reliability of the ,control instructions and data transmission.

[0090] See also Figure 1 ,Furthermore, the IoT device control module communicates with the IoT ,device and sends control instructions to the device according to the control ,decision made by the control and decision module, so as to remotely control and ,manage the IoT devices to improve the management efficiency and response speed of the IoT ,device.

[0091] When used specifically, first, the distributed data acquisition module collects data from various IoT devices, including sensor data and device status, and transmits the data to the control system. Then, the data normalization module standardizes and uniformly processes the collected data to ensure data consistency and comparability. The data storage and management module stores the data in a suitable storage medium and organizes, indexes, controls access, and manages backup and recovery operations to ensure data consistency, integrity, and availability. Secondly, the big data mining module proposes a big data mining algorithm based on fuzzy clustering to analyze and mine the effective information of the stored data to discover the patterns, trends, and correlations hidden behind the big data, and provide support for subsequent decision-making. To provide support, the control and decision-making module includes a big data analysis control unit and a big data decision execution unit. The big data analysis control unit proposes a big data analysis control algorithm based on reinforcement learning to conduct in-depth analysis of the effective information of the big data extracted by the big data mining module, identify the status, abnormal conditions and trend changes of the Internet of Things devices, and provide real-time control suggestions. Then, the big data decision execution unit receives the results from the analysis control unit and formulates corresponding control strategies and decision plans. The secure communication module ensures the security of internal and external communications of the system, including data encryption, identity authentication, and access control. Finally, the Internet of Things device control module communicates and controls the Internet of Things devices, including sending direct control instructions and receiving device status feedback.

[0092] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A big data-based IoT device control system, comprising a distributed data acquisition module, a data normalization module, a data storage and management module, a big data mining module, a control and decision module, a secure communication module, and an IoT device control module, characterized in that: The distributed data acquisition module is used to collect data from various IoT devices, including sensor data and device status, and transmit the data to the control system. The data normalization module is used to standardize the format of the collected data and uniformly process it to ensure the consistency and comparability of the data. The data storage and management module is used to store the data in a suitable storage medium and organize, index, access control, backup and restore the data to ensure the consistency, integrity and availability of the data. The big data mining module proposes a big data mining algorithm based on fuzzy clustering to analyze and mine the effective information of the stored data to discover the patterns, trends and correlations hidden behind the big data and provide support for subsequent decision-making. , the control and decision-making module includes a big data analysis control unit and a big data decision execution unit. The big data analysis control unit proposes a big data analysis control algorithm based on reinforcement learning to conduct in-depth analysis of the effective information of the big data extracted by the big data mining module, identify the state, abnormal conditions and trend changes of the Internet of Things devices, and provide real-time control suggestions. The big data decision execution unit is used to receive the results from the analysis control unit and formulate corresponding control strategies and decision plans. The secure communication module is used to ensure the security of internal and external communications of the system, including data encryption, identity authentication, and access control. The Internet of Things device control module is used to communicate and control with the Internet of Things devices, including sending direct control instructions and receiving device status feedback; The big data mining module proposes a big data mining algorithm based on fuzzy clustering, deeply mines the effective information of the stored big data, and reveals the potential information and regularity in the data through pattern recognition, association analysis and anomaly detection, so as to provide support and insight for decision-making; First, the original large data set X processed by the data normalization module is divided into data blocks of equal size according to the IoT devices. Each data block is an independent unit and can be processed independently, which helps to improve the efficiency and speed of the algorithm. The formula for data block division is as follows: X={X1,X2,...,X M }={X i |i=1,...,M} Among them, X represents the original large data set, X1, X2, ..., X M Represented as a data block, X i It represents the i-th data block after division, i represents the index of the number of data blocks, and M represents the total number of data blocks. Secondly, incremental fuzzy clustering is performed on each data block. The specific formula is as follows: Among them, q sv Represented as data point x s With cluster center c v The degree of membership between sv The larger the value, the stronger the data point x s More likely to belong to cluster center c v The cluster where d sv Represented as data point x s With cluster center c v The distance between ls Represented as data point x s With cluster center c l The distance between them, η is represented by the fuzzy coefficient, which is used to control the fuzziness of the membership, s is represented by the index of the data point, v is represented by the index of the cluster center, l∈v is represented by the index of the lth cluster center, min(·) is represented by the minimum value function, and the data point x s Represented as the data in each data block, the specific formula is: X i ={x s |s=1,...,S} Among them, S represents the total number of data points, and the cluster center c v The specific initialization formula is as follows: Among them, λ s Represented as data point x s The weight is used to consider the importance of the data point, τ is expressed as a constant coefficient, and then the specific formula for updating the cluster center position is as follows: Among them, c v (t) represents the vth cluster center after updating at time t, Represented as data block X i The distance between the cluster center c at the previous moment t-1 v (t-1) the most recent data point, Represented as data points With the initial cluster center c v The distance between them, the calculation formula of the transfer point coefficient of each IoT device is as follows: in, It is represented as the transfer point coefficient between the ith IoT device and the jth IoT device at time t, j∈i is the index of the jth IoT device, Represented as data block X at time t i and X j The inverse matrix of the kernel function values ​​between Represented as data block X at time t i and X j The kernel function value between It is expressed as the transfer point coefficient between the i-th IoT device and the j-th IoT device at time t-1. By mapping the new cluster center to the data block space, the optimal transfer point can be selected. The specific calculation formula is expressed as: Among them, Φ(c v (t),j) represents the new cluster center of the jth class at time t, Φ(X i ) represents the data block X i All data points in the data are mapped to a function in a high-dimensional space. The design of the entire algorithm fully utilizes the information and value hidden in big data to achieve intelligent control and optimization of IoT devices. Through intelligent data mining, the algorithm can discover potential patterns and laws in the data and provide a more accurate reference for decision-making. The multi-transmission point selection method further improves the accuracy and efficiency of data analysis, ensuring the full utilization and value mining of big data.

2. According to the big data-based IoT device control system of claim 1, it is characterized by: The distributed data acquisition module works simultaneously through multiple data acquisition nodes to efficiently collect data from various IoT devices and transmits the data to the data storage and management module in real time, ensuring the efficiency and real-time performance of data acquisition and avoiding single point failures affecting the overall system operation.

3. The IoT device control system based on big data according to claim 1 is characterized in that: The data normalization module ensures the consistency and comparability of data during storage and processing by normalizing data formats and standardizing data fields, reduces the complexity and error rate of data processing, and ensures that data of different types and formats can be processed in a unified and standardized manner for subsequent data storage, management and analysis.

4. The IoT device control system based on big data according to claim 3 is characterized in that: The data storage and management module securely stores the collected data and effectively manages the big data through data indexing, authority control, and data backup functions, thereby ensuring the effectiveness and reliability of data organization, management, and maintenance.

5. The IoT device control system based on big data according to claim 1 is characterized in that: The control and decision-making module includes a big data analysis control unit and a big data decision execution unit. The big data analysis control unit proposes a big data analysis control algorithm based on reinforcement learning, and performs real-time analysis on the effective information of the big data extracted by the big data mining module to identify patterns, trends and abnormal situations in the data, and timely update control suggestions to support the system's intelligent control and decision-making of IoT devices; the big data decision execution unit formulates specific control strategies based on the big data analysis results, and sends control instructions to the security communication module to realize remote control and management of IoT devices.

6. The IoT device control system based on big data according to claim 5 is characterized in that: First, determine the state space and executable action space of the IoT device. The specific expression formulas of the state space and action space are: Among them, S represents the state space of IoT devices, Represents the IoT device status, represents the mth state of the ith IoT device, i represents the number index of IoT devices, I represents the total number of IoT devices, m represents the state index of each IoT device, A represents the action space of IoT devices, Represents the actions of IoT devices, It is represented as the nth action of the ith IoT device, and n represents the action index of each IoT device. By defining the state space and action space, the control problem of the IoT device is transformed into a reinforcement learning problem, which provides a basis for the subsequent establishment of strategies. Then, the reward function is constructed as a function of the state of the system and the action performed at each time step. The specific formula is expressed as: in, It is expressed as the state at time t Next action The reward function is It is expressed as the state at time t Next action The energy efficiency reward value obtained after , w1 represents the weight coefficient of the energy efficiency reward part, It is expressed as the state at time t Next action The performance optimization reward value obtained after , w2 is expressed as the weight coefficient of the performance optimization reward part, It is expressed as the state at time t Next action The communication efficiency reward value obtained after that, w3 is expressed as the weight coefficient of the communication efficiency reward part. Then, the Actor-Critic structure is used to build a reinforcement learning model, where the Actor model is used to learn the strategy of selecting actions in a given state, and the Critic model is used to calculate the expected return of performing an action in a given state. The specific formulas of the Actor model and the Critic model are expressed as follows: Among them, O Actor (θ) represents the optimization objective function of the Actor model, θ represents the parameter of the Actor model, is expressed as the expected function, min(·) is expressed as the minimum function, It is expressed as a given state at time t under the Actor model parameter θ Select action The probability of It is expressed as a given state at time t-1 under the Actor model parameter θ Select action The probability of Expressed as the advantage function at time t, it is used to measure the execution of actions Compared with the average level, clip is represented by the clipping function, which is used to limit the range of the ratio, ∈ is represented by the clipping parameter, which is used to control the range of clipping, O Critic (μ) is the optimization objective function of the Critic model, which is used to update the parameters of the Critic model. Expressed as the Critic model for the state at time t The action value estimation function is represented by |·|, and μ is represented by the parameter of the Critic model. By combining the reinforcement learning architecture with multiple IoT devices to share learning experience and model parameters, the system can learn and control multiple devices more robustly. This cooperative learning method can not only improve the robustness of the system, but also achieve adaptive control of different devices.

7. The IoT device control system based on big data according to claim 1, characterized in that: The secure communication module encrypts and authenticates data through a secure communication protocol to prevent unauthorized access, theft and tampering of the data, protect the confidentiality and integrity of the communication data, and ensure the security and reliability of control instructions and data transmission.

8. The IoT device control system based on big data according to claim 1, characterized in that: The IoT device control module communicates with the IoT device, sends control instructions to the device according to the control decision made by the control and decision module, and remotely controls and manages the IoT device to improve the management efficiency and response speed of the IoT device.

Images