Internet of Things sensor equipment data acquisition system and method

Abstract

The invention discloses an Internet of Things sensor equipment data acquisition system and method. The system is composed of a main controller, a sensor module group and a dynamic sampling frequency adjustment module. The main controller is responsible for executing a dynamic sampling frequency adjustment strategy and protocol adaptation, the sensor module group is connected with the main controller through a standardized interface and adopts modular design, and each module is provided with a plugging interface and a multi-protocol support module. The dynamic sampling frequency adjustment module is integrated with a data acquisition and change rate calculation unit, a sampling frequency adjustment unit and a machine learning optimization unit, and can dynamically adjust the sampling frequency according to the change of environmental parameters. The method comprises the steps of system initialization configuration, multi-protocol data acquisition and conversion, dynamic sampling frequency adjustment and machine learning optimization recommendation. Through a dynamic sampling frequency adjustment mechanism, the system can sense the change trend of environmental parameters in real time, and dynamically optimize the sampling frequency based on a machine learning algorithm, thereby reducing redundant data.

Description

Technical Field

[0001] The present invention relates to the field of sensor data acquisition applications, and in particular to a data acquisition system and method for sensor equipment in the Internet of Things. Background Art

[0002] With the rapid development of IoT technology, more and more devices are collecting and monitoring data through sensor networks. As the core components of IoT devices, sensors are responsible for real-time monitoring of environmental parameters (such as temperature, humidity, pressure, gas concentration, etc.) and transmitting the data to a central monitoring system.

[0003] During data collection, traditional sensor systems are often affected by sampling frequency, sensor accuracy, and environmental interference, resulting in data delays or errors, making them unable to meet the needs of high-precision, real-time monitoring. With the continuous expansion of IoT applications, sensor networks need to support more device types and protocols. Existing technologies lack system scalability and compatibility, making it difficult to meet diverse needs and the operational requirements of sensor data collection applications. Therefore, a data collection system and method for IoT sensor devices is proposed. Summary of the Invention

[0004] The present invention provides the following technical solution: an Internet of Things sensor device data acquisition system, comprising: A main controller, a sensor module group, and a dynamic sampling frequency adjustment module. The main controller is used to execute the dynamic sampling frequency adjustment strategy and protocol adaptation. The sensor module group is connected to the main controller through a standardized interface and is used to collect environmental parameters. A dynamic sampling frequency adjustment module is used to dynamically adjust the sampling frequency according to changes in environmental parameters. The dynamic sampling frequency adjustment module integrates a data acquisition and change rate calculation unit, a sampling frequency adjustment unit, and a machine learning optimization unit; The sensor module group adopts a modular design. Each module in the sensor module group is equipped with a plug-in interface and a multi-protocol support module. The sampling frequency adjustment formula of the dynamic sampling frequency adjustment module is: The f in the sampling frequency adjustment formula current is the current sampling frequency, a and b are the ratio of the rate of change and the ratio of the cumulative change of adjacent windows respectively, a threshold and b threshold is the preset threshold, α and β are weight coefficients; The machine learning optimization unit uses the LSTM model to predict future data changes and recommend sampling frequency. The formula for the recommended sampling frequency is: In the formula of the recommended sampling frequency, ΔV_pred is the data change amplitude, Δt is the time interval, and ∈ is the fault tolerance coefficient.

[0005] The present invention provides a method for collecting data from an Internet of Things sensor device, based on the above-mentioned Internet of Things sensor device data collection system, comprising the following steps: S1 system initialization configuration: First, configure the main controller parameters, including the dynamic sampling frequency adjustment strategy and protocol adaptation engine. Next, initialize the sensor module group, detect the sensor type, interface status, and multi-protocol support module, and set the initial parameters of the dynamic sampling frequency adjustment module. Finally, load the pre-trained LSTM model and online learning function parameters in the machine learning optimization unit. S2 multi-protocol data acquisition and conversion: The sensor module group is connected through a standardized interface. The sensor module group then collects environmental parameters such as temperature, humidity, light and gas concentration in real time. The main controller then converts the collected data into a format that can be processed by the main controller through a protocol adaptation engine. S3 dynamically adjusts the sampling frequency: The data acquisition and change rate calculation unit uses a sliding window mechanism to analyze environmental parameters in real time, calculate the maximum, minimum, average and standard deviation of the data, and use the weighted moving average method to calculate the change rate based on the data statistics of adjacent time windows. The sampling frequency adjustment unit then dynamically adjusts the sampling frequency based on the change rate ratio and the cumulative change ratio using a fuzzy control algorithm. S4 machine learning optimization recommendations: The machine learning optimization unit uses the LSTM model to predict future data changes. The input includes current environmental parameters, historical environmental parameter sequences, and system status information. S5 energy consumption awareness mode switch: The energy management unit dynamically adjusts the operating mode of the sensor module according to the system load and battery power, and simultaneously applies the support vector machine regression algorithm to predict the remaining battery life based on historical discharge data. Then, based on the prediction results and system requirements, it dynamically adjusts the sensor module's supply voltage and sampling frequency.

[0006] Preferably, a protocol adaptation engine is integrated inside the main controller, and the protocol adaptation engine achieves protocol compatibility by dynamically loading a protocol parsing library, and the protocol parsing library includes a protocol header parsing module, a data frame checking module and a data conversion module.

[0007] Preferably, the sensor module group includes a temperature sensor module, a humidity sensor module, a light sensor module and a gas sensor module, and the sensor modules inside the sensor module group are all equipped with dual redundant plug-in interfaces.

[0008] Preferably, the data acquisition and change rate calculation unit adopts a sliding window mechanism to perform real-time analysis of environmental parameters. The sliding window mechanism is used to divide the continuously collected data into multiple time windows of equal length. The data acquisition and change rate calculation unit calculates the maximum value, minimum value, average value and standard deviation of the data in each time window, and calculates the change rate based on the data statistics of adjacent time windows. The calculation of the change rate is performed using a weighted moving average method.

[0009] Preferably, the sampling frequency adjustment unit adopts a fuzzy control algorithm to dynamically adjust the sampling frequency. The fuzzy control algorithm takes the ratio of the change rate and the ratio of the cumulative change as input variables, and takes the sampling frequency adjustment amount as the output variable. The membership function of the fuzzy control algorithm is customized according to the sensor characteristics and application scenarios, and the sampling frequency adjustment unit is internally provided with an amplitude adjustment limit function.

[0010] Preferably, the input data of the LSTM model includes current environmental parameters, historical environmental parameter sequences and system status information; the output of the LSTM model is a recommended sampling frequency sequence for a period of time in the future; the machine learning optimization unit is internally provided with an online learning function, which is used to perform incremental training on the LSTM model based on real-time collected environmental parameter data and system feedback; the online learning function adopts a small-batch gradient descent method, and the batch size and learning rate of the small-batch gradient descent method are dynamically adjusted according to the system resource occupancy.

[0011] Preferably, the dynamic sampling frequency adjustment module also has an energy consumption management unit integrated therein, and the energy consumption management unit is used to dynamically adjust the working mode of the sensor module according to the system load and the battery power. The working modes include normal working mode, low power consumption mode and sleep mode. The energy consumption management unit is provided with a battery power prediction function therein, and the battery power prediction function is used to predict the remaining battery life based on historical discharge data. The battery power prediction function adopts a support vector machine regression algorithm.

[0012] Preferably, in step S3, the adjustment of the sampling frequency must satisfy the following constraints: fmin≤fnew≤min(fmax_sensor,fmax_system) In the constraints, fmax_sensor is the maximum frequency supported by the sensor hardware, and fmax_system is the upper limit of the system processing capability.

[0013] Preferably, in step S4, when the machine learning optimization unit is working, the online learning function performs incremental training on the LSTM model based on the real-time collected environmental parameter data and system feedback, and adopts the small batch gradient descent method.

[0014] In summary, compared with the prior art, the present invention provides a data acquisition system and method for IoT sensor devices, which has the following beneficial effects: 1. The present invention uses a dynamic sampling frequency adjustment mechanism to perceive the changing trends of environmental parameters in real time and dynamically optimizes the sampling frequency based on a machine learning algorithm. This allows the system to automatically reduce the sampling frequency when the environment changes slowly, significantly reducing the generation of redundant data. When the environmental parameters fluctuate rapidly, the sampling frequency can be quickly increased to ensure that key change information is not missed. Compared with traditional fixed sampling systems, this reduces the data storage and transmission load and improves the overall operating efficiency of the system. At the same time, the LSTM prediction model integrated in the system can predict the changing trends of environmental parameters in advance by learning historical data patterns, realize forward-looking sampling frequency adjustment, and avoid the lag in response speed of traditional methods. 2. The present invention simplifies the system deployment and maintenance process by adopting a standardized interface design and plug-and-play mechanism, and the introduction of a multi-protocol support module further enhances the compatibility of the system, enabling it to seamlessly connect to various sensor devices using different communication protocols, effectively solving the device heterogeneity problem that is prevalent in the Internet of Things environment. At the same time, this modular architecture makes it possible to connect the corresponding sensor module when new monitoring parameters need to be added without having to modify the entire system, thereby reserving sufficient space for future functional expansion of the system. Users can add new sensor types or communication protocols at any time according to business development needs, ensuring that the system can continue to adapt to changing application requirements and enhancing the compatibility and scalability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a structural schematic diagram of the present invention.

[0016] Figure 2 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0018] See also Figure 1The present invention provides a technical solution, a data acquisition system for an Internet of Things sensor device, comprising: a main controller, a sensor module group, and a dynamic sampling frequency adjustment module, wherein the main controller is used to execute a dynamic sampling frequency adjustment strategy and protocol adaptation, the sensor module group is connected to the main controller via a standardized interface, and the sensor module group is used to collect environmental parameters; A dynamic sampling frequency adjustment module is used to dynamically adjust the sampling frequency according to changes in environmental parameters. The dynamic sampling frequency adjustment module integrates a data acquisition and change rate calculation unit, a sampling frequency adjustment unit, and a machine learning optimization unit; The sensor module group adopts a modular design. Each module in the sensor module group is equipped with a plug-in interface and a multi-protocol support module. The sampling frequency adjustment formula of the dynamic sampling frequency adjustment module is: f in the sampling frequency adjustment formula current is the current sampling frequency, a and b are the ratio of the rate of change and the ratio of the cumulative change of adjacent windows respectively, a threshold and b threshold is the preset threshold, α and β are weight coefficients; The machine learning optimization unit uses the LSTM model to predict future data changes and recommend sampling frequencies. The formula for recommending sampling frequencies is: In the formula for the recommended sampling frequency, ΔV_pred is the data change amplitude, Δt is the time interval, and ∈ is the fault tolerance coefficient; The main controller is internally integrated with a protocol adaptation engine, which achieves protocol compatibility by dynamically loading a protocol parsing library. The protocol parsing library includes a protocol header parsing module, a data frame check module, and a data conversion module. The sensor module group includes a temperature sensor module, a humidity sensor module, a light sensor module and a gas sensor module. The sensor modules inside the sensor module group are all equipped with dual redundant plug-in interfaces; The data acquisition and change rate calculation unit uses a sliding window mechanism to perform real-time analysis of environmental parameters. The sliding window mechanism is used to divide the continuously collected data into multiple time windows of equal length. The data acquisition and change rate calculation unit calculates the maximum, minimum, average and standard deviation of the data in each time window, and calculates the change rate based on the data statistics of adjacent time windows. The change rate is calculated using the weighted moving average method. The sampling frequency adjustment unit uses a fuzzy control algorithm to dynamically adjust the sampling frequency. The fuzzy control algorithm uses the ratio of the rate of change and the ratio of the cumulative change as input variables and the sampling frequency adjustment amount as output variable. The membership function of the fuzzy control algorithm is customized according to the sensor characteristics and application scenarios. The sampling frequency adjustment unit is internally provided with an amplitude adjustment limit function. The input data of the LSTM model includes current environmental parameters, historical environmental parameter sequences, and system status information. The output of the LSTM model is a sequence of recommended sampling frequencies for a period of time in the future. The machine learning optimization unit is internally equipped with an online learning function, which is used to incrementally train the LSTM model based on real-time collected environmental parameter data and system feedback. The online learning function uses a small-batch gradient descent method, and the batch size and learning rate of the small-batch gradient descent method are dynamically adjusted according to system resource usage. The dynamic sampling frequency adjustment module also integrates an energy consumption management unit, which is used to dynamically adjust the working mode of the sensor module according to the system load and battery power. The working modes include normal working mode, low power mode and sleep mode. The energy consumption management unit is equipped with a battery power prediction function, which is used to predict the remaining battery life based on historical discharge data. The battery power prediction function adopts a support vector machine regression algorithm.

[0019] See also Figure 2 The present invention provides a data collection method for an Internet of Things sensor device, based on the above-mentioned data collection system for an Internet of Things sensor device, comprising the following steps: S1 system initialization configuration: First, configure the main controller parameters, including the dynamic sampling frequency adjustment strategy and protocol adaptation engine. Next, initialize the sensor module group, detect the sensor type, interface status, and multi-protocol support module, and set the initial parameters of the dynamic sampling frequency adjustment module. Finally, load the pre-trained LSTM model and online learning function parameters in the machine learning optimization unit. S2 multi-protocol data acquisition and conversion: The sensor module group is connected through a standardized interface. The sensor module group then collects environmental parameters such as temperature, humidity, light and gas concentration in real time. The main controller then converts the collected data into a format that can be processed by the main controller through a protocol adaptation engine. S3 dynamically adjusts the sampling frequency: The data acquisition and change rate calculation unit uses a sliding window mechanism to analyze environmental parameters in real time, calculate the maximum, minimum, average, and standard deviation of the data, and calculate the change rate using the weighted moving average method based on the data statistics of adjacent time windows. The sampling frequency adjustment unit then dynamically adjusts the sampling frequency based on the change rate ratio and the cumulative change ratio using a fuzzy control algorithm. The adjustment of the sampling frequency must meet the following constraints: fmin≤fnew≤min(fmax_sensor,fmax_system) In the constraints, fmax_sensor is the maximum frequency supported by the sensor hardware, and fmax_system is the upper limit of the system processing capability. S4 machine learning optimization recommendations: The machine learning optimization unit uses an LSTM model to predict future data changes. The input includes current environmental parameters, historical environmental parameter sequences, and system status information. When the machine learning optimization unit is working, the online learning function can incrementally train the LSTM model based on the real-time collected environmental parameter data and system feedback using a small-batch gradient descent method. The specific process of the above method is as follows: Data preprocessing and feature engineering: When the online learning function is activated, the environmental parameter data collected in real time is first standardized. The system automatically identifies and removes abnormal data points, and fills missing values ​​with the weighted average of adjacent time windows. The preprocessed data is organized according to time series to form a feature vector containing current environmental parameters, historical change trends, and system status information. The feature engineering module dynamically extracts the statistical characteristics of the data, including features such as the mean, variance, and range within the sliding window, and calculates the difference index from the previous time window. These features will be organized into a time series data format suitable for LSTM model processing; Incremental training data preparation: The system maintains a fixed-size training data buffer and uses a first-in-first-out strategy to manage data samples collected in real time. When new data arrives, the system combines it with historical data in the buffer into training batches. Each training batch contains a data sequence of multiple consecutive time steps to ensure continuity in the time dimension. The system dynamically adjusts the batch size based on current resource usage to ensure training effectiveness while avoiding memory overload. The training data is divided into input features and target values ​​in chronological order, where the target value is the actual observed change pattern of environmental parameters; Model incremental training execution: The LSTM model is updated online using mini-batch gradient descent. During training, the system monitors the model's performance on the validation set in real time and dynamically adjusts the learning rate. After each iteration, the model parameters are carefully updated, with the update amplitude limited by system-preset constraints to prevent sudden changes in model performance due to the characteristics of a single batch of data. The training process uses an early stopping mechanism, automatically terminating the current training cycle when the model's performance improvement over multiple consecutive batches falls below a threshold. The system records model performance indicators after each incremental training to evaluate the effectiveness of online learning. Model validation and deployment: After completing incremental training, the system will evaluate the model performance on an independent validation dataset. The validation process simulates prediction tasks in real scenarios, and the evaluation indicators include prediction accuracy, response latency, and resource consumption. The model version that passes the validation will be marked as a candidate version and A / B tested with the model in the current production environment. During the test, the two versions of the model will run in parallel, and the system will compare their prediction results under the same conditions. The model with better performance will be automatically deployed to the production environment, while the previous version of the model will be retained as a rollback backup; Feedback loop and continuous optimization: The system establishes a complete prediction-execution-feedback closed-loop mechanism. Each time the sampling frequency is adjusted based on the model's recommended value, the system records the quality indicators of the actual data collected and compares and analyzes them with the predicted values. This feedback data is incorporated into the data preparation phase of the next round of training, forming a positive feedback loop of continuous optimization. The system also regularly conducts full training on long-term accumulated data to eliminate model bias that may be caused by incremental training. The entire online learning process is monitored by the resource management module to ensure that model training does not affect the system's real-time data collection function; S5 energy consumption awareness mode switch: The energy management unit dynamically adjusts the operating mode of the sensor module based on the system load and battery power level, and simultaneously applies a support vector machine regression algorithm to predict the remaining battery life based on historical discharge data. It then dynamically adjusts the sensor module's supply voltage and sampling frequency based on the prediction results and system requirements. The specific process of the above method is as follows: Real-time system status monitoring: The energy management unit continuously monitors the operating status of the sensor network, collecting key indicators such as the current remaining battery charge, instantaneous power consumption, and system task load. The monitoring module records the operating current and voltage of each sensor module at fixed intervals to establish a complete energy consumption characteristic data set. The system also maintains a historical discharge database that stores detailed records of past charge and discharge cycles, including parameters such as discharge rate and the impact of ambient temperature on battery performance. After preprocessing, the monitoring data is classified and labeled and fed into the predictive analysis module. Battery life prediction and analysis phase: A support vector machine regression algorithm is used to process historical discharge data. The prediction model first standardizes the input features to eliminate the influence of different dimensions. The system automatically selects the most predictive feature combination, including battery cycle count, average depth of discharge, ambient temperature change pattern, etc. The prediction process takes battery aging characteristics into account and establishes specialized feature engineering for battery performance degradation at different stages of use. The prediction results not only include an estimate of the remaining battery life, but also output confidence intervals and possible abnormality warnings, providing a multi-dimensional reference for subsequent decision-making; Dynamic working mode decision-making stage: Based on the battery life prediction results and current system requirements, the energy management unit executes a multi-level decision-making process. First, the priority of key monitoring tasks is evaluated to ensure that the collection of important parameters is not affected; then, based on the remaining power prediction value, the optimal configuration is selected between normal working mode, low power mode and sleep mode. Fuzzy logic is introduced into the decision-making process to handle the mode switching problem under boundary conditions and avoid additional energy consumption caused by frequent switching. The system will generate multiple candidate configuration schemes, simulate and evaluate the expected battery life of each scheme, and finally select the optimal solution that balances performance and battery life; Power supply parameter fine-tuning phase: After determining the operating mode, the system further optimizes the power supply parameters of each sensor module. Dynamic voltage regulation technology is used to allocate appropriate operating voltages to sensors of different importance levels. High-priority sensors maintain standard power supply, while secondary sensors operate at reduced voltage. Simultaneously, the system coordinates with the sampling frequency adjustment module to moderately reduce the acquisition frequency of non-critical parameters. The adjustment process uses a gradual change strategy to avoid measurement errors caused by sudden changes in voltage and frequency. The system continuously monitors the actual energy-saving effects after adjustment to accumulate empirical data for subsequent optimization. Closed-Loop Feedback and Adaptive Optimization Phase: After each adjustment, the system establishes a comprehensive performance tracking mechanism to record deviations from the expected target energy consumption. This data is fed back into the prediction model for online correction of prediction parameters. The system regularly retrains the support vector machine model to adapt to changes in battery characteristics due to aging. When abnormal discharge is detected, a special optimization process is automatically triggered to extend battery life in critical situations by reducing non-essential functions. This entire adjustment process forms a closed-loop control, ensuring that the energy management strategy continuously adapts to changes in the actual operating environment.

[0020] This solution uses a dynamic sampling frequency adjustment mechanism to perceive the changing trends of environmental parameters in real time, and dynamically optimizes the sampling frequency based on a machine learning algorithm. This allows the system to automatically reduce the sampling frequency when the environment changes slowly, significantly reducing the generation of redundant data. When environmental parameters fluctuate rapidly, the sampling frequency can be quickly increased to ensure that key change information is not missed. Compared with traditional fixed sampling systems, this reduces the data storage and transmission load and improves the overall operating efficiency of the system. At the same time, the system's integrated LSTM prediction model can predict the changing trends of environmental parameters in advance by learning historical data patterns, realize forward-looking sampling frequency adjustment, and avoid the lag in response speed of traditional methods.

[0021] This solution simplifies the system deployment and maintenance process by adopting standardized interface design and plug-and-play mechanism. The introduction of multi-protocol support modules further enhances the compatibility of the system, enabling it to seamlessly connect various sensor devices using different communication protocols, effectively solving the device heterogeneity problem that is prevalent in the Internet of Things environment. At the same time, this modular architecture makes it possible to connect the corresponding sensor module when new monitoring parameters need to be added without having to transform the entire system, thereby reserving sufficient space for future functional expansion of the system. Users can add new sensor types or communication protocols at any time according to business development needs, ensuring that the system can continue to adapt to changing application needs and enhancing the system's compatibility and scalability.

[0022] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0023] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An Internet of Things sensor device data acquisition system, characterized in that: include: A main controller, a sensor module group, and a dynamic sampling frequency adjustment module. The main controller is used to execute the dynamic sampling frequency adjustment strategy and protocol adaptation. The sensor module group is connected to the main controller through a standardized interface and is used to collect environmental parameters. A dynamic sampling frequency adjustment module is used to dynamically adjust the sampling frequency according to changes in environmental parameters. The dynamic sampling frequency adjustment module integrates a data acquisition and change rate calculation unit, a sampling frequency adjustment unit, and a machine learning optimization unit; The sensor module group adopts a modular design. Each module in the sensor module group is equipped with a plug-in interface and a multi-protocol support module. The sampling frequency adjustment formula of the dynamic sampling frequency adjustment module is: The f in the sampling frequency adjustment formula current is the current sampling frequency, a and b are the ratio of the rate of change and the ratio of the cumulative change of adjacent windows respectively, a threshold and b threshold is the preset threshold, α and β are weight coefficients; The machine learning optimization unit uses the LSTM model to predict future data changes and recommend sampling frequency. The formula for the recommended sampling frequency is: In the formula of the recommended sampling frequency, ΔV_pred is the data change amplitude, Δt is the time interval, and ∈ is the fault tolerance coefficient.

2. The IoT sensor device data acquisition system according to claim 1, characterized in that: The main controller is internally integrated with a protocol adaptation engine, which realizes protocol compatibility by dynamically loading a protocol analysis library. The protocol analysis library includes a protocol header analysis module, a data frame check module and a data conversion module.

3. The IoT sensor device data acquisition system according to claim 1, characterized in that: The sensor module group includes a temperature sensor module, a humidity sensor module, a light sensor module and a gas sensor module. The sensor modules inside the sensor module group are all equipped with dual redundant plug-in interfaces.

4. The IoT sensor device data acquisition system according to claim 1, characterized in that: The data acquisition and change rate calculation unit uses a sliding window mechanism to perform real-time analysis of environmental parameters. The sliding window mechanism is used to divide the continuously collected data into multiple time windows of equal length. The data acquisition and change rate calculation unit calculates the maximum value, minimum value, average value and standard deviation of the data in each time window, and calculates the change rate based on the data statistics of adjacent time windows. The calculation of the change rate is performed using the weighted moving average method.

5. The IoT sensor device data acquisition system according to claim 1, characterized in that: The sampling frequency adjustment unit uses a fuzzy control algorithm to dynamically adjust the sampling frequency. The fuzzy control algorithm uses the change rate ratio and the cumulative change ratio as input variables and the sampling frequency adjustment amount as output variable. The membership function of the fuzzy control algorithm is customized according to the sensor characteristics and application scenarios. The sampling frequency adjustment unit is internally provided with an amplitude adjustment limit function.

6. The IoT sensor device data acquisition system according to claim 1, characterized in that: The input data of the LSTM model includes current environmental parameters, historical environmental parameter sequences and system status information. The output of the LSTM model is a recommended sampling frequency sequence for a period of time in the future. The machine learning optimization unit is internally provided with an online learning function, which is used to perform incremental training on the LSTM model based on real-time collected environmental parameter data and system feedback. The online learning function adopts a small-batch gradient descent method, and the batch size and learning rate of the small-batch gradient descent method are dynamically adjusted according to the system resource occupancy.

7. The IoT sensor device data acquisition system according to claim 1, characterized in that: The dynamic sampling frequency adjustment module also has an integrated energy consumption management unit, which is used to dynamically adjust the working mode of the sensor module according to the system load and battery power. The working modes include normal working mode, low power mode and sleep mode. The energy consumption management unit is provided with a battery power prediction function, which is used to predict the remaining battery life based on historical discharge data. The battery power prediction function adopts a support vector machine regression algorithm.

8. A method for collecting data from an Internet of Things sensor device, based on the Internet of Things sensor device data collection system according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1 system initialization configuration: First, configure the main controller parameters, including the dynamic sampling frequency adjustment strategy and protocol adaptation engine. Next, initialize the sensor module group, detect the sensor type, interface status, and multi-protocol support module, and set the initial parameters of the dynamic sampling frequency adjustment module. Finally, load the pre-trained LSTM model and online learning function parameters in the machine learning optimization unit. S2 multi-protocol data acquisition and conversion: The sensor module group is connected through a standardized interface, and then the sensor module group collects environmental parameters in real time. The main controller then converts the collected data into a format that can be processed by the main controller through the protocol adaptation engine. S3 dynamically adjusts the sampling frequency: The data acquisition and change rate calculation unit uses a sliding window mechanism to analyze environmental parameters in real time, calculate the maximum, minimum, average and standard deviation of the data, and use the weighted moving average method to calculate the change rate based on the data statistics of adjacent time windows. The sampling frequency adjustment unit then dynamically adjusts the sampling frequency based on the change rate ratio and the cumulative change ratio using a fuzzy control algorithm. S4 machine learning optimization recommendations: The machine learning optimization unit uses the LSTM model to predict future data changes. The input includes current environmental parameters, historical environmental parameter sequences, and system status information. S5 energy consumption awareness mode switch: The energy management unit dynamically adjusts the operating mode of the sensor module according to the system load and battery power, and simultaneously applies the support vector machine regression algorithm to predict the remaining battery life based on historical discharge data. Then, based on the prediction results and system requirements, it dynamically adjusts the sensor module's supply voltage and sampling frequency.

9. The method for collecting data from an Internet of Things sensor device according to claim 8, wherein: In step S3, the adjustment of the sampling frequency satisfies the following constraints: fmin≤fnew≤min(fmax_sensor,fmax_system) In the constraints, fmax_sensor is the maximum frequency supported by the sensor hardware, and fmax_system is the upper limit of the system processing capability.

10. The method for collecting data from an Internet of Things sensor device according to claim 8, wherein: In step S4, when the machine learning optimization unit is working, the online learning function can perform incremental training on the LSTM model based on the real-time collected environmental parameter data and system feedback, and adopt the small batch gradient descent method.

Images