Internet of Things data processing method and Internet of Things system

Abstract

The invention discloses an Internet of Things data processing method and an Internet of Things system, and the Internet of Things data processing method comprises the following steps: S1, data collection; s2, data transmission; s3, data preprocessing; s4, data analysis; and S5, decision execution, and the Internet of Things data processing system comprises a sensing layer, a network layer, a processing layer and an application layer. According to the method, the adaptive sampling technology and the multi-link transmission technology are combined with a dynamic bandwidth allocation strategy, and the algorithm based on machine learning and deep learning is adopted, so that the system is endowed with powerful data processing capability, and a more comprehensive basis is provided for decision making by combining an association rule mining algorithm, and the accuracy and the real-time performance of data processing are improved; and the intelligent development of the application of the Internet of Things is powerfully promoted.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a method for processing IoT data and an IoT system. Background Technology

[0002] With the widespread application of IoT devices, the amount of data is growing explosively. Existing IoT systems face numerous challenges in data processing. On the one hand, the sheer volume of data overwhelms traditional processing architectures, leading to low processing efficiency. For example, when smart factory equipment generates massive amounts of data, significant processing delays occur, impacting production scheduling. On the other hand, real-time performance is difficult to guarantee. In smart transportation scenarios, untimely traffic data processing hinders effective congestion relief.

[0003] Moreover, traditional processing methods are simple, limited to data collection, transmission, and storage, and lack in-depth analysis. Fixed sampling frequencies easily collect redundant data, single transmission links are unstable when the network is congested, data preprocessing is crude, and data analysis methods are simple, making it difficult to extract data value and failing to meet the ever-expanding application needs of the Internet of Things.

[0004] Accordingly, this application proposes a method for processing Internet of Things (IoT) data and an IoT system. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for processing Internet of Things (IoT) data and an IoT system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for processing Internet of Things (IoT) data includes the following steps:

[0008] S1. Data Acquisition: IoT devices collect various types of data through multiple sensors, including physical quantity data such as temperature, humidity, and pressure, as well as device status data. At the same time, adaptive sampling technology is adopted to dynamically adjust the sampling interval according to the frequency of data change. The sampling frequency is increased for data with drastic changes and decreased for relatively stable data to reduce the collection of invalid data and improve data acquisition efficiency.

[0009] S2. Data transmission: Utilizing multi-link transmission technology, the collected data is transmitted through multiple different network links, including wired and wireless networks. During transmission, the data transmission volume is dynamically allocated based on the real-time network status to ensure that the data can be transmitted to the data processing center quickly and stably. At the same time, data encryption and verification technologies are used to ensure the security and integrity of data transmission.

[0010] S3. Data preprocessing: Clean the transmitted data to remove noise and duplicate data; use machine learning-based anomaly detection algorithms to identify and mark abnormal data, providing a basis for subsequent processing.

[0011] S4. Data Analysis: Utilize deep learning algorithms to perform in-depth analysis on preprocessed data; construct time series prediction models to predict future trends of the data; and simultaneously, employ association rule mining algorithms to analyze the relationships between different data points and extract valuable information.

[0012] S5. Decision Execution: Based on the results of data analysis, the system automatically generates decision instructions and sends them to the corresponding IoT devices for execution; at the same time, the decision results are fed back to the user, providing visualized reports and suggestions.

[0013] Preferably, in step S1, a sensor device management system is established to monitor the working status of the sensors in real time, including power and signal strength. When a sensor malfunctions or becomes abnormal, an alarm is issued in a timely manner and automatic switching or repair is performed.

[0014] Adaptive sampling technology incorporates artificial intelligence algorithms to predict data trends and adjust sampling intervals in advance based on historical data and real-time operating conditions of the device.

[0015] Preferably, in step S2, the data encryption and verification technology adopts an advanced encryption algorithm; at the same time, multiple verification technologies such as CRC cyclic redundancy check and hash check are used to ensure the integrity of the data.

[0016] Preferably, in step S3, the anomaly detection algorithm combines support vector machine and clustering analysis algorithms with the neural network model for anomaly detection; by integrating multiple algorithms, the accuracy and reliability of anomaly detection are improved.

[0017] Preferably, in step S4, the deep learning algorithm selects a suitable deep learning algorithm based on the characteristics of the data and the analysis objectives; at the same time, the parameters of the algorithm are optimized to improve the accuracy and generalization ability of the model.

[0018] Time series forecasting models incorporate external factors to improve their accuracy; at the same time, ensemble learning methods are used to fuse the results of multiple time series forecasting models, further enhancing forecasting performance.

[0019] Preferably, in step S5, the decision instruction formulates detailed decision instruction generation rules based on different data analysis results and business needs.

[0020] An Internet of Things (IoT) data processing system includes:

[0021] The perception layer, as the foundation of the Internet of Things system, consists of a large number of sensor nodes and smart devices. These devices not only cover common temperature, humidity and pressure sensors, but also innovatively integrate biosensors, gas sensors and other sensors to meet the diverse data collection needs of different application scenarios.

[0022] The network layer adopts an advanced hybrid network architecture that integrates 5G, Wi-Fi 6, and low-power wide area network (LPWAN) technologies. 5G and Wi-Fi 6 are responsible for high-speed, low-latency data transmission, meeting the needs of applications with high real-time requirements. LPWAN technology is suitable for large amounts of low-power, long-distance data transmission.

[0023] The processing layer consists of edge computing nodes and cloud computing centers working together. Edge computing nodes are deployed close to the data source and have real-time data processing capabilities. They can perform preliminary analysis and filtering of the collected data, reducing the burden on the cloud computing center. The cloud computing center is responsible for in-depth mining and analysis of massive amounts of data, using powerful computing resources to run complex machine learning and deep learning algorithms to achieve comprehensive insights and decision support from the data.

[0024] The application layer provides users with a wide variety of application services, covering multiple fields such as industry, agriculture, healthcare, and smart homes; by developing various applications and user interfaces, the processed data is presented to users in an intuitive and easy-to-understand way.

[0025] Preferably, the sensing layer has self-calibration and self-diagnosis functions to ensure that the sensor nodes and smart devices operate stably for a long time and that the data acquisition is accurate and reliable; at the same time, it has adaptive adjustment capabilities, which can automatically adjust the acquisition parameters according to environmental changes to ensure the validity of the data.

[0026] Preferably, the network layer incorporates blockchain technology to encrypt and authenticate the data transmission process, ensuring data security and immutability, and further enhancing the trust mechanism of the Internet of Things system.

[0027] Preferably, the edge computing node has a data caching function, which can temporarily store data when the network fails and upload it after the network is restored, ensuring data integrity; the cloud computing center has a powerful fault tolerance and backup mechanism to ensure data security, and supports multi-tenant mode to provide independent computing and storage resources for different users.

[0028] The present invention has the following beneficial effects:

[0029] 1. Highly efficient data acquisition: Adaptive sampling technology dynamically adjusts the sampling interval based on the frequency of data changes, avoiding over-collection of stable data and reducing the amount of data collected and transmission costs. From the perspective of improving data processing efficiency, adaptive sampling technology reduces the collection of invalid data. During the data transmission phase, the amount of data to be transmitted is significantly reduced, accelerating data transmission speed in the network and alleviating network burden. In the data preprocessing and analysis phases, the workload of removing noise and duplicate data is reduced, and machine learning-based anomaly detection algorithms can operate more efficiently, quickly and accurately identifying abnormal data, thereby comprehensively improving the efficiency of the entire data processing workflow.

[0030] 2. Stable and fast data transmission: Multi-link transmission technology combined with a dynamic bandwidth allocation strategy intelligently selects the optimal transmission path based on real-time network conditions. In complex network environments, the system automatically allocates more data traffic to the wired network, ensuring rapid data transmission. Simultaneously, data encryption and verification technologies employ advanced encryption algorithms and verification mechanisms to prevent data from being stolen, tampered with, or lost during transmission, ensuring the security and integrity of data transmission and providing a reliable data foundation for subsequent data processing.

[0031] 3. Precise and intelligent data processing: Based on machine learning and deep learning algorithms, the system possesses powerful data processing capabilities. In the data preprocessing stage, machine learning algorithms accurately identify and remove noise and duplicate data, improving data purity. In the data analysis stage, deep learning models can delve into the potential patterns and regularities within the data. Association rule mining algorithms help discover the inherent connections between different data points, providing a more comprehensive basis for decision-making. The system automatically generates and executes decision instructions based on the data analysis results, achieving intelligent decision-making, improving the accuracy and real-time performance of data processing, and powerfully promoting the intelligent development of IoT applications. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall process of an IoT data processing method proposed in this invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0034] Example 1:

[0035] Data acquisition was conducted at the large-scale vegetable planting base, where various sensors were deployed. Temperature and humidity sensors employed capacitive humidity sensing elements and thermistor temperature sensing elements, achieving measurement accuracy of ±2% RH (relative humidity) and ±0.3℃. Soil pH sensors utilized ion-selective electrode technology, controlling the measurement error to within ±0.1 pH. Light sensors used photodiodes sensitive to different spectra, with a light intensity measurement error range of ±5 Lux. The UAV-borne multispectral image sensor met the requirements for flight altitude H (meters) and image resolution H (pixels / square meter). (K is a constant that depends on the sensor performance), thereby obtaining high-resolution macroscopic data on vegetation cover and crop growth status;

[0036] An adaptive sampling strategy is adopted, assuming a temperature change rate. Humidity change rate Where T n H n T represents the current temperature and humidity. n-1 H n-1 The temperature and humidity at the previous moment are represented by ΔT, where ΔT is the time interval (in hours). The initial sampling frequency of the sensor is f0, which is once every 30 minutes (0.5 hours). When |ΔT| > 2℃ / hour or |ΔH| > 5% / hour, the sampling frequency is increased to once every 5 minutes. Once per hour. For example, if T n-1 =25℃, T n =28℃, ΔT = 1 hour, ΔT = 3℃ / hour, the conditions are met, and the sampling frequency is adjusted to f1. Through this strategy, redundant data is reduced by a certain percentage. According to statistics, it is approximately 70%, and N0 and N1 are the amounts of data collected before and after adaptive sampling, respectively.

[0037] For data transmission, the base adopts a hybrid network architecture of LoRa and 4G. The transmission distance d (kilometers) of the LoRa network signal is related to factors such as transmit power P (dBm), antenna gain G (dBi), and receiver sensitivity S (dBm), with an approximate relationship as follows: (k1 is an environment-related constant), suitable for large areas with stable signals; the 4G network transmission rate V (Mbps) is related to the signal strength I (dBm) and bandwidth B (MHz) as V = k2 × I × B (k2 is a constant), used for transmission in complex environments;

[0038] Regarding the dynamic transmission strategy, let the signal strength be I. When I > -80dBm, LoRa will be used preferentially for data transmission. Let the total data volume be D. When the LoRa network bandwidth is B... LoRa With 4G network bandwidth B 4G satisfy At that time, the amount of data transmitted by LoRa is D LoRa=D × 0.7, 4G data transmission volume D 4G =D × 0.3. With this strategy, the data transmission success rate... It can reach over 99%, P success N total These represent the amount of data successfully transmitted and the total amount of data transmitted, respectively. The data is encrypted using the AES-256 algorithm to ensure security.

[0039] Data preprocessing, based on machine learning-based anomaly detection algorithms, involves calculating the Mahalanobis distance between data points and the mean μ of the normal feature distribution. To identify anomalies, where x represents a data point and ∑ is the covariance matrix. When d M Exceeding the set threshold d th (e.g., d is obtained through training with a large amount of historical data) th =3) is considered abnormal data. For example, for temperature data point x = [26, 27, 25], it is known that... If d M =3.5>d th If it does not, it is considered abnormal. The accuracy rate of abnormal data identification... Up to 95%, N correct N detected These represent the amount of correctly identified anomalous data and the total amount of detected anomalous data, respectively.

[0040] After data cleaning to remove duplicate data, features closely related to crop growth are extracted, such as temperature change trends. The duration of illumination is calculated by recording the start and end times of illumination using sensors, while the change in illumination intensity is calculated by determining the standard deviation of illumination intensity over different time periods. To measure, L i The light intensity at each moment, denoted as average illumination intensity, and m as the number of sampling points.

[0041] Data analysis utilizes deep learning algorithms to construct crop growth models. The model inputs include temperature (T), humidity (H), light intensity (L), and soil pH. The output is either crop yield (Y) or the probability of pest and disease occurrence (P). disease The model is trained using a large amount of historical data, and the parameter θ is adjusted to improve the model's prediction values. or Mean square error of the actual value (production forecast) or (Prediction of disease and pest probability) is minimized, where n is the number of samples;

[0042] For example, by analyzing historical and real-time environmental data, an outbreak of pests and diseases can be predicted 7 days in advance, with a prediction accuracy rate of [missing information]. Up to 80%, N true N totalpredictThese represent the number of correctly predicted pest and disease occurrences and the total number of predictions, respectively. The model continuously learns from new data to optimize parameters and improve accuracy.

[0043] Decision execution, when the probability P of pests and diseases occurs disease When the pesticide application rate exceeds 70%, the system automatically activates the drone to perform pesticide spraying. The drone's flight speed v (m / s), spraying width w (m), and pesticide spraying rate Q (liters / m²) satisfy the following conditions: Where t is the spraying time, Q0 is the pesticide concentration, and s is the area of ​​the pest and disease zone;

[0044] Simultaneously, early warning information and prevention and control suggestions are sent to farmers. Based on the type of pests and diseases and historical prevention and control experience, decision tree algorithms and other methods are used to determine the prevention and control methods and pesticide dosages. Through precise decision-making and implementation, the pest and disease loss rate is reduced. Reduced by 50%, crop yield increase rate 20%, Y before Y after These represent the output before and after implementation, respectively.

[0045] Example 2:

[0046] In smart home systems, data acquisition follows a precise placement principle for various devices. Temperature and humidity sensors are positioned away from heat sources, cold sources, and ventilation openings to ensure measurement accuracy, achieving ±2% RH (relative humidity) and ±0.3℃. Human infrared sensors are installed at room entrances and in key activity areas, with detection range angle α (degrees), installation height h (meters), and coverage radius r (meters) meeting specific requirements. The relationship between the shooting angle β (degrees), resolution R (pixels), and monitored area S (square meters) of a smart camera installed in public areas such as the living room and hallway is as follows: k1 is a constant;

[0047] The system employs a dynamic sampling mechanism. Let ΔM be the change in human activity detected by the human infrared sensor. When ΔM > 0, i.e., when someone is detected entering the room, the sampling frequency of the human infrared sensor changes from every 10 minutes... The system switches from hourly monitoring to real-time monitoring (sampled once per second, i.e., f1 = 1 time / second), and the smart camera starts recording video. When no one is detected in the room for time t... idle Exceeding the set time t threshold (e.g., 30 minutes, i.e., t) threshold When the frame rate of the smart camera drops from F1 to F2 (e.g., from 30 frames per second to 10 frames per second) after 0.5 hours, the human infrared sensor resumes low-frequency sampling. This mechanism reduces the proportion of invalid data. N0 and N1 represent the amount of data collected before and after adopting the dynamic sampling mechanism, respectively, which effectively improves the targeting of data collection.

[0048] In data transmission, the transmission method is intelligently selected based on the data volume and device characteristics. Let the data volume of the smart camera be D. camera The temperature and humidity sensor data volume is D sensor When D camera When the signal strength is >100KB and the device is close to the router (signal strength I > -60dBm), transmission is via Wi-Fi; when D... sensor When the data size is less than 10KB and the device supports Bluetooth, transmission is performed via Bluetooth. The data transmission amount D and the transmission time t satisfy the following condition: V represents the transmission rate, such as the Wi-Fi transmission rate V0. Wi-Fi Or Bluetooth transmission rate V Bluetooth For example, if the data volume D of the smart camera... camera =200KB, V Wi-Fi =50Mbps (converted to KB / s is 50×1024÷8=6400KB / s), then the transmission time

[0049] When the Wi-Fi network is congested, i.e., the network latency L > 50ms, according to the relationship between video resolution and data volume D = k2 × R 2 (k2 is a constant, R2 is the resolution), reduce the video resolution. For example, reduce the resolution from 1920×1080 (R1 = 1920×1080 pixels) to 1280×720 (R2 = 1280×720 pixels), the data size increases from D1 = k2 × (1920×1080). 2 The result becomes D2 = k2 × (1280 × 720). 2 This reduces data transmission volume. Data verification uses CRC-16 checksum, which has a high accuracy rate. According to statistical data, the transmission error rate is less than 0.1%, N correct N total These represent the amount of data that was verified correctly and the total amount of data transmitted, respectively.

[0050] Data preprocessing involves using machine learning algorithms to build a data feature model to identify outlier data. Let the normal range for temperature and humidity data be T. min ≤T≤T max H min ≤H≤H max When data T is detected new or H new Data outside this range is marked as abnormal. For example, if the normal temperature range T... min =20℃, T max =30℃, T was detectednew If the temperature is 35℃, then the temperature data is considered abnormal. Abnormal data identification accuracy. This accuracy can be improved by training the model, N correct N detected These represent the amount of correctly identified anomalous data and the total amount of detected anomalous data, respectively.

[0051] A home environment comfort model is constructed using deep learning algorithms. The comfort index C is categorized into the relationship between factors such as temperature (T), humidity (H), and light intensity (L) as C = k3 × T + k4 × H + k5 × L + ..., where k3, k4, k5, etc., are weighting coefficients determined through training the model with extensive historical data. For example, when T = 25℃, H = 50% RH, L = 500 Lux, and k3 = 0.5, k4 = 0.3, k5 = 0.2, C = 0.5 × 25 + 0.3 × 50 + 0.2 × 500 = 12.5 + 15 + 100 = 127.5. Based on the model output, air conditioning, curtains, and other equipment are automatically adjusted to keep C within the comfort range.

[0052] In terms of decision execution, the system features energy-saving control and comfort adjustment functions. When it detects that no one is home and electrical appliances are still running, it automatically sends a power-off command. Let the total power consumption of the electrical appliances be P. total The running time is t, and the power consumption is E = P. total ×t. Through intelligent decision-making and execution, the household's electricity consumption decreased by a certain percentage. According to statistics, it is 15%, E before E after These represent electricity consumption before and after implementation. For example, electricity consumption E in the month before implementation. before =300 kWh, monthly electricity consumption E after implementation after =255 degrees, then

[0053] When the indoor temperature T > 28℃, the air conditioner will automatically turn on for cooling. Let the cooling capacity of the air conditioner be Q. cool Q is related to factors such as indoor area S and temperature difference ΔT. cool = k6 × S × ΔT, where k6 is a constant. For example, if the indoor area S = 30 square meters, the temperature difference ΔT = 3℃, and k6 = 50, then Q cool = 50 × 30 × 3 = 4500 (units are determined based on actual cooling capacity). When light intensity L > L threshold (e.g. L) threshold When the value reaches 1000 Lux, the curtains will automatically close to enhance living comfort.

[0054] Example 3:

[0055] In logistics warehouses, various sensors perform different functions in data acquisition. Cargo weight sensors are installed on load-bearing parts of shelves and transport vehicles, with a measurement accuracy of ±0.5 kg. Position sensors utilize GPS or indoor positioning technology, achieving an outdoor positioning accuracy of ±5 meters and an indoor accuracy of ±1 meter. Temperature and humidity sensors are evenly distributed throughout the warehouse, with temperature and humidity measurement accuracies of ±2% RH (relative humidity) and ±0.3℃, respectively.

[0056] The system employs an adaptive sampling and adjustment strategy. Assume the rate of change in cargo weight is... Among them W n W represents the current weight of the cargo. n-1 Let |ΔW| represent the weight of the goods at the previous moment, and Δt represent the time interval (in hours). When |ΔW| > 10 kg, the sampling frequency of the weight sensor is increased from once per hour (f0 = 1) to once per minute (f0 = 1). Once per hour. For example, if the weight W of the goods at the previous moment... n-1 =50kg, current time W n =65kg, Δt=1 hour, at this time If the condition |ΔW|>10kg is met, the sampling frequency is adjusted to f1. This strategy reduces the proportion of invalid data. According to statistics, it is approximately 60%, and N0 and N1 are the amounts of data collected before and after adaptive sampling, respectively.

[0057] Data transmission relies on wired Ethernet and 5G networks. Inside the warehouse, wired Ethernet ensures stable data transmission rates, with a transmission rate of V... ethernet It is closely related to the type of network cable and the performance of the network equipment. For example, when using Cat 6e network cable and the network equipment supports gigabit transmission, V ethernet It can reach 1000Mbps (i.e., 1000 × 1024 ÷ 8 = 128000 KB / s). When the freight vehicle is en route, the 5G network undertakes the task of transmitting data such as vehicle location and cargo status in real time, and its transmission rate V 5G The relationship between V and signal strength I (dBm) and bandwidth B (MHz) is as follows: 5G = k1 × I × B (k1 is a constant). For example, when the signal strength I = -700dBm, the bandwidth B = 100MHz, and k1 = 0.1, V 5G =0.1×(-70)×100=-700Mbps (the actual transmission rate is positive; this is just an example calculation);

[0058] To ensure the real-time transmission of critical logistics data, transmission priorities are dynamically allocated based on the network status parameter S (which comprehensively considers factors such as bandwidth and latency). S is calculated by weighting parameters such as network bandwidth B and latency L, using the following formula: Among them B currentGiven the current network bandwidth, B max For maximum network bandwidth, L current For the current network latency, L max The maximum acceptable network latency is defined as S. When S > 0.8, the network is considered congested. In this state, the transmission priority of critical logistics data (such as vehicle location and cargo safety status) is set to 1, and the priority of non-critical data is set to 0.5. Simultaneously, data is encrypted using SSL / TLS encryption protocol to ensure data security.

[0059] In the data preprocessing stage, anomaly detection algorithms are used to identify abnormal data. Let W be the normal range for cargo weight. min ≤W≤W max When the weight W of the goods is detected new Data exceeding this range is considered abnormal. For example, if W... min =20kg, W max =100kg, W was detected new If the weight is 120kg, then this data is considered abnormal. Abnormal data identification accuracy. This accuracy can be continuously improved by training the model, N. correct To correctly identify the amount of anomalous data, N detected This represents the total amount of abnormal data detected.

[0060] A logistics transportation model is constructed using deep learning algorithms. The model input includes data such as vehicle speed v (km / h), road condition information (e.g., congestion level C, where C ranges from 0 to 1, with 0 indicating smooth traffic and 1 indicating severe congestion), and cargo weight W. The output is the cargo transportation time T. trans (hours), vehicle failure probability P fault The model is trained using a large amount of historical data to improve its predictions. Mean absolute error from actual value (Transportation time forecast) or (Fault probability prediction) is minimized, where n is the sample size. For example, by analyzing vehicle driving data and road condition information, a possible vehicle fault can be predicted 2 hours in advance, achieving a prediction accuracy of [missing value]. Up to 75%, N true To predict the correct number of failures, N totalpredict This represents the total number of predictions.

[0061] Decision execution, when the vehicle failure probability P fault When the efficiency is greater than 80%, the system automatically initiates emergency dispatch. The arrival time T of the equipment vehicle... arrive The relationship between distance d (km) and travel speed v (km / h) is as follows: For example, if the backup vehicle is 50 kilometers away from the disabled vehicle (d = 50 km) and traveling at a speed of 60 km / h (v = 60 km / h), then... Within hours. Simultaneously, the system sends maintenance suggestions and information on nearby repair stations to the driver. Through intelligent decision-making and execution, the on-time delivery rate of logistics transportation is improved. Statistics show an increase of 25%, N on-time N represents the number of shipments that arrive on time. total Total number of transports;

[0062] Furthermore, logistics processes are optimized using logistics transportation models. For example, by analyzing historical transportation data and real-time road conditions, algorithms such as Dijkstra's algorithm are employed to optimize transportation routes. Let the starting point be A, the destination be B, and the weights of each edge in the graph represent road travel time. Algorithms calculate the shortest path, thereby reducing transportation time. (Cargo damage rate) Statistics show a 30% reduction, N damaged N represents the quantity of damaged goods. total This represents the total quantity of goods.

[0063] It should be noted that the comparison included three implementation examples: smart agriculture, smart home, and smart logistics, as well as two comparative examples: traditional fixed sampling single-link transmission and simple algorithm processing. From the five dimensions of data acquisition, transmission, preprocessing, analysis, and decision execution, it can be seen that:

[0064] Examples 1 and 3 employ adaptive sampling, which adjusts the sampling frequency according to environmental changes, significantly reducing redundant data by approximately 70% and 60% respectively, ensuring that no critical data is missed. Example 2 uses dynamic sampling, changing the equipment's operating mode according to personnel activities, thus improving the targeting of data collection. In contrast, Comparative Examples 1 and 2 use fixed sampling, which cannot adjust with environmental changes, leading to the easy omission of critical data and the generation of a large amount of invalid data.

[0065] Regarding data transmission, Example 1 utilizes a hybrid LoRa and 4G network, flexibly allocating transmission volume based on signal strength and network conditions, achieving a data transmission success rate exceeding 99%. Example 2 intelligently selects Wi-Fi or Bluetooth transmission based on data volume and device characteristics, resulting in a data transmission error rate of less than 0.1%. Example 3 uses wired Ethernet within the warehouse and a 5G network during transportation, dynamically allocating transmission priorities based on network status parameters to ensure real-time transmission of critical data. In contrast, Examples 1 and 2 rely on a single network, making data transmission prone to interruption or severe delays when the signal is unstable or congested, thus failing to adapt to complex network environments.

[0066] In the data preprocessing stage, Example 1 achieved an anomaly detection accuracy of 95%. While Examples 2 and 3 did not specify concrete figures, they demonstrated that the anomaly detection accuracy significantly improved after training. Comparative Example 1 lacked an effective anomaly detection algorithm, making it difficult to accurately identify anomalous data; Comparative Example 2 only performed simple data cleaning and could not leverage machine learning and other techniques for in-depth analysis.

[0067] In terms of data analysis, Example 1 can predict pest and disease outbreaks 7 days in advance with 80% accuracy and build a crop growth model; Example 2 can build a home environment comfort model, realizing in-depth utilization of environmental data; Example 3 can predict vehicle malfunctions 2 hours in advance with 75% accuracy and build a logistics transportation model. In contrast, Comparative Example 1 can only perform simple statistical analysis and is difficult to achieve accurate prediction; Comparative Example 2 cannot build a high-precision model at all and cannot provide strong support for decision-making.

[0068] In terms of decision-making execution, Example 1 reduced pest and disease losses by 50% and increased crop yield by 20%; Example 2 achieved a 15% reduction in electricity consumption and effectively improved living comfort; Example 3 improved logistics on-time delivery by 25% and reduced cargo damage rate by 30%. In contrast, Comparative Example 1 lacked precision and timeliness in decision-making, and Comparative Example 2 relied on human experience and lacked intelligent automation, resulting in system operating efficiency and benefits far lower than the Examples 1.

[0069] In summary, the smart agriculture, smart home, and smart logistics implementation examples demonstrate significant advantages in all dimensions of data collection, transmission, preprocessing, analysis, and decision execution. Compared to the two comparative examples, they can perform data processing and decision execution more efficiently, stably, and accurately.

[0070] It should be noted that Examples 1, 2, and 3, and Comparative Examples 1 and 2 are compared in terms of device compatibility, scalability, system response time, energy consumption reduction rate, data storage requirement reduction rate, and security protection level, as shown in Table 1:

[0071] Table 1: Comparison of Items between Examples and Comparative Examples

[0072]

[0073]

[0074] As shown in Table 1, in terms of device compatibility, all three implementation examples (smart agriculture, smart home, and smart logistics) performed well, scoring 4 points, indicating that they can be well compatible with various devices and meet the needs of multiple devices working together in different scenarios. However, the two comparative examples only scored 2 points, which means that their device compatibility is poor and there may be many limitations in device access and collaboration.

[0075] In terms of scalability, all three embodiments demonstrated strong capabilities, scoring 4 points, indicating that they can easily add new devices and expand new functions as business grows or functional requirements increase. In contrast, the traditional fixed-sampling single-link transmission ratio and the simple algorithm processing ratio showed lower scalability, scoring only 2 points, which may lead to difficulties in adapting to subsequent system upgrades and business expansion.

[0076] Regarding system response time, the smart agriculture and smart logistics implementations both have response times of less than 5 seconds, while the smart home implementation is even less than 3 seconds, demonstrating the advantage of rapid response and the ability to react promptly to environmental changes or user commands. In contrast, the two comparative implementations have response times greater than 10 seconds or even 15 seconds. This could lead to serious latency issues in scenarios with high real-time requirements, affecting the normal operation of the system and the user experience.

[0077] In terms of energy consumption reduction, the smart agriculture implementation showed a 20% reduction, the smart home implementation a 15% reduction, and the smart logistics implementation a 10% reduction, all achieving a certain degree of energy saving. In stark contrast, the two comparative implementations showed no improvement in energy consumption reduction, with a reduction rate of 0%, meaning they were far less energy efficient than the three implementations.

[0078] In terms of reducing data storage requirements, the smart agriculture implementation achieved a 70% reduction, the smart logistics implementation a 60% reduction, and the smart home implementation a 50% reduction. This is attributed to their respective adoption of adaptive sampling and dynamic sampling strategies, which significantly reduced the storage of redundant data. In contrast, the comparative implementation, lacking these optimization mechanisms, did not reduce its data storage requirements and instead consumed more storage resources.

[0079] In terms of security protection levels, the smart agriculture and smart logistics implementation examples scored 4 points, while the smart home implementation example scored 3 points, indicating a high level of overall security protection capabilities that effectively ensure data security and stable system operation. The two comparative examples, however, scored only 2 points, which is relatively low and suggests significant vulnerabilities in data security and resistance to external risks.

[0080] In summary, the smart agriculture, smart home, and smart logistics implementation examples demonstrate outstanding overall performance across all projects, showing significant advantages compared to the two comparative examples. They are better suited to the development needs of modern IoT applications, achieving multiple goals of performance optimization and security protection while ensuring efficient and stable system operation.

[0081] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for processing Internet of Things (IoT) data, characterized in that, Includes the following steps: S1. Data Acquisition: IoT devices collect various types of data through multiple sensors, including physical quantity data such as temperature, humidity, and pressure, as well as device status data. At the same time, adaptive sampling technology is adopted to dynamically adjust the sampling interval according to the frequency of data change. The sampling frequency is increased for data with drastic changes and decreased for relatively stable data to reduce the collection of invalid data and improve data acquisition efficiency. S2. Data transmission: Using multi-link transmission technology, the collected data is transmitted through multiple different network links, including wired networks and wireless networks. During transmission, the data transmission volume is dynamically allocated based on the real-time network status to ensure that the data can be transmitted to the data processing center quickly and stably; at the same time, data encryption and verification technologies are used to ensure the security and integrity of data transmission. S3. Data preprocessing: Clean the transmitted data to remove noise and duplicate data; use machine learning-based anomaly detection algorithms to identify and mark abnormal data, providing a basis for subsequent processing. S4. Data Analysis: Utilize deep learning algorithms to perform in-depth analysis on preprocessed data; construct time series prediction models to predict future trends of the data; and simultaneously, employ association rule mining algorithms to analyze the relationships between different data points and extract valuable information. S5. Decision Execution: Based on the results of data analysis, the system automatically generates decision instructions and sends them to the corresponding IoT devices for execution; at the same time, the decision results are fed back to the user, providing visualized reports and suggestions.

2. The method for processing Internet of Things (IoT) data according to claim 1, characterized in that, In step S1, a sensor device management system is established to monitor the working status of the sensors in real time, including power and signal strength. When a sensor malfunctions or becomes abnormal, an alarm is issued in a timely manner and automatic switching or repair is performed. Adaptive sampling technology incorporates artificial intelligence algorithms to predict data trends and adjust sampling intervals in advance based on historical data and real-time operating conditions of the device.

3. The method for processing Internet of Things (IoT) data according to claim 1, characterized in that, In step S2, the data encryption and verification technology employs advanced encryption algorithms; at the same time, it uses multiple verification technologies such as CRC cyclic redundancy check and hash check to ensure data integrity.

4. The method for processing Internet of Things (IoT) data according to claim 1, characterized in that, In step S3, the anomaly detection algorithm combines neural network models with support vector machines and clustering analysis to perform anomaly detection; by integrating multiple algorithms, the accuracy and reliability of anomaly detection are improved.

5. The method for processing Internet of Things (IoT) data according to claim 1, characterized in that, In step S4, the deep learning algorithm selects a suitable deep learning algorithm based on the characteristics of the data and the analysis objectives; at the same time, the parameters of the algorithm are optimized to improve the accuracy and generalization ability of the model. Time series forecasting models incorporate external factors to improve their accuracy; at the same time, ensemble learning methods are used to fuse the results of multiple time series forecasting models, further enhancing forecasting performance.

6. The method for processing Internet of Things (IoT) data according to claim 1, characterized in that, In step S5, the decision-making instructions formulate detailed rules for generating decision-making instructions based on different data analysis results and business needs.

7. A data processing system for the Internet of Things (IoT), characterized in that, include: The perception layer, as the foundation of the Internet of Things system, consists of a large number of sensor nodes and smart devices. These devices not only cover common temperature, humidity and pressure sensors, but also innovatively integrate biosensors, gas sensors and other sensors to meet the diverse data collection needs of different application scenarios. The network layer adopts an advanced hybrid network architecture that integrates 5G, Wi-Fi 6, and low-power wide area network (LPWAN) technologies. 5G and Wi-Fi 6 are responsible for high-speed, low-latency data transmission, meeting the needs of applications with high real-time requirements. LPWAN technology is suitable for large amounts of low-power, long-distance data transmission. The processing layer consists of edge computing nodes and cloud computing centers working together. Edge computing nodes are deployed close to the data source and have real-time data processing capabilities. They can perform preliminary analysis and screening of the collected data, reducing the burden on the cloud computing center. The cloud computing center is responsible for in-depth mining and analysis of massive amounts of data, using powerful computing resources to run complex machine learning and deep learning algorithms to achieve comprehensive insights and decision support from the data. The application layer provides users with a wide variety of application services, covering multiple fields such as industry, agriculture, healthcare, and smart homes; by developing various applications and user interfaces, the processed data is presented to users in an intuitive and easy-to-understand way.

8. The IoT data processing system according to claim 7, characterized in that, The sensing layer has self-calibration and self-diagnosis functions to ensure that sensor nodes and smart devices operate stably for a long time and that data acquisition is accurate and reliable. At the same time, it has adaptive adjustment capabilities, which can automatically adjust the acquisition parameters according to environmental changes to ensure the validity of the data.

9. The IoT data processing system according to claim 7, characterized in that, The network layer incorporates blockchain technology to encrypt and authenticate the data transmission process, ensuring data security and immutability, and further enhancing the trust mechanism of the Internet of Things system.

10. The IoT data processing system according to claim 7, characterized in that, The edge computing nodes have data caching capabilities, allowing them to temporarily store data during network failures and upload it again once the network is restored, ensuring data integrity. The cloud computing center has robust fault tolerance and backup mechanisms to ensure data security, while also supporting multi-tenant mode to provide independent computing and storage resources for different users.

Images