Cabinet environment monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of digital information transmission, and particularly discloses a cabinet environment monitoring method and system based on the Internet of Things. The system comprises multiple types of sensor nodes, an edge computing gateway and a cloud monitoring platform, the reporting priority and interval are adjusted in a self-adaptive mode through a dynamic reporting strategy module according to the data state and the change rate, and data transmission optimization is achieved in combination with a time division multiple access mechanism of a communication resource scheduling module. According to the invention, the network load and the system energy consumption are reduced, and the real-time performance and the reliability of the key alarm are improved at the same time.

Description

Technical Field

[0001] This invention belongs to the field of digital information transmission technology, specifically relating to a cabinet environment monitoring method and system based on the Internet of Things. Background Technology

[0002] In the field of information technology infrastructure, server racks, as the basic units that house critical hardware such as servers and network equipment, are crucial for ensuring the reliable operation of the entire system due to the stability and security of their internal environment. Server rack environment monitoring technology aims to perceive the physical state parameters inside the rack in real time, providing data support for equipment maintenance and fault early warning.

[0003] Among them, the IoT-based rack environment monitoring system achieves comprehensive perception of the internal environment of the rack by deploying various sensors. The system collects environmental data through devices such as temperature and humidity sensors, vibration sensors, and smoke detectors, and uploads the data to the monitoring center for analysis and processing using network transmission technology, thereby achieving real-time monitoring of the rack environment status.

[0004] In existing technologies, various sensors typically use a fixed-frequency data reporting mechanism. This mechanism generates a large amount of redundant data when network conditions are poor or the cabinet status is stable, which not only occupies valuable network bandwidth resources but also increases the pressure on cloud data processing systems.

[0005] Meanwhile, heterogeneous sensor data streams may compete for communication resources during transmission, leading to delays in the transmission of critical alarm information and failing to meet the timeliness requirements of real-time monitoring. Furthermore, the fixed-frequency data reporting mode also results in high power consumption, which is detrimental to the long-term stable operation of the system. These issues collectively constitute the technical challenges that urgently need to be addressed in the field of rack environment monitoring. Summary of the Invention

[0006] The technical problem this invention aims to solve is to overcome the shortcomings of existing rack environment monitoring systems, such as high network bandwidth consumption, heavy cloud data processing pressure, transmission delay of critical alarm information, and high system power consumption caused by the fixed-frequency data reporting mechanism. To achieve the above objective, this invention provides a rack environment monitoring method and system based on the Internet of Things (IoT).

[0007] This invention provides a cabinet environment monitoring method and system based on the Internet of Things (IoT), including multiple types of sensor nodes deployed inside the cabinet, an edge computing gateway communicating with the sensor nodes, and a monitoring and analysis platform located in the cloud. The multiple types of sensor nodes include at least temperature and humidity sensors, vibration sensors, and smoke detectors. The edge computing gateway has a built-in data acquisition and preprocessing module, a dynamic reporting strategy decision-making module, and a communication resource scheduling module. The monitoring and analysis platform includes a data receiving and storage unit, a status assessment and prediction model, and a strategy parameter distribution interface.

[0008] The data acquisition and preprocessing module is responsible for receiving raw monitoring data from various types of sensor nodes and filtering and standardizing the data. The dynamic reporting strategy decision module calculates the data reporting priority and reporting time interval for each sensor node in real time based on the preprocessed data stream and preset state assessment rules. The communication resource scheduling module allocates transmission time slots and controls the data transmission behavior of the sensor nodes according to the priority sequence and time interval parameters output by the dynamic reporting strategy decision module.

[0009] Furthermore, the operating mechanism of the dynamic reporting strategy decision module includes the following steps: First, a baseline state threshold range and an alarm state threshold are set for each type of sensor data; second, the absolute value of the rate of change of the current data relative to the previous reported data is calculated in real time; then, based on whether the current data exceeds the baseline state threshold range and the magnitude of the absolute value of the rate of change, the reporting priority level of the sensor node is determined through a multi-dimensional decision table. The reporting priority level is divided into 3 levels, with level 1 being the highest priority and level 3 being the lowest priority. Finally, based on the determined reporting priority level, the preset priority-reporting time interval mapping relationship is queried to obtain the specific time interval for the next data reporting. This mapping relationship is set so that the time interval corresponding to level 1 priority is 2s, the time interval corresponding to level 2 priority is 30s, and the time interval corresponding to level 3 priority is 300s.

[0010] Furthermore, when any sensor data exceeds its alarm status threshold, the dynamic reporting strategy decision module immediately forces the reporting priority of that node to level 1 and generates a highest priority alarm data packet; when the communication resource scheduling module detects the highest priority alarm data packet, it immediately interrupts the current non-urgent data transmission and prioritizes allocating a transmission channel for the alarm data packet.

[0011] Furthermore, the communication resource scheduling module employs a time-division multiple access mechanism to schedule data transmission. This module maintains a dynamic time slot allocation table with a time slot length of 10ms. Each sensor node is woken up and sends data within a specific time slot according to its assigned reporting time interval and priority. For nodes with priority level 1, the system reserves dedicated time slots to ensure low-latency transmission. Nodes with priority levels 2 and 3 share the remaining time slots and coordinate the transmission timing through a lightweight conflict avoidance algorithm.

[0012] Furthermore, the state assessment and prediction model runs on the monitoring and analysis platform. This model is a deep learning model based on a long short-term memory network. The model uses historical monitoring data to predict the future state of the cabinet environment, with a prediction period of 5 minutes. The output of the model includes the predicted values ​​of each sensor data within the next 5 minutes and a state stability index. The monitoring and analysis platform sends the state stability index to the edge computing gateway through the policy parameter distribution interface. The dynamic reporting policy decision module of the edge computing gateway receives the state stability index and adaptively adjusts the baseline state threshold range accordingly. When the state stability index is higher than 0.8, the baseline state threshold range is appropriately widened. When the state stability index is less than 0.5, the baseline state threshold range is narrowed.

[0013] Furthermore, the multi-type sensor nodes adopt a working mode that combines event-driven and timed wake-up. During non-reporting times, the sensor nodes are in a low-power sleep state. The sensor nodes are only woken up and perform data acquisition and reporting operations when one of the following three conditions is met: Condition 1, the next reporting time point specified by the edge computing gateway is reached; Condition 2, a sudden change occurs in the physical quantity detected by the sensor itself, and the change magnitude is greater than a preset local change threshold; Condition 3, an emergency polling instruction is received from the edge computing gateway.

[0014] Furthermore, the monitoring and analysis platform also has a policy feedback optimization loop; this loop continuously collects data on network bandwidth utilization, alarm response time, and node energy consumption during actual operation; based on this operational data, the platform periodically optimizes the multi-dimensional decision table and the mapping relationship between priority and reporting time interval in the dynamic reporting policy decision module through a reinforcement learning algorithm, with an optimization cycle of 24 hours.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention introduces a dynamic reporting strategy decision-making mechanism, adaptively adjusting the reporting frequency and priority based on the real-time status and changing trends of sensor data, completely changing the fixed-frequency reporting mode. When the rack environment is stable, the system automatically reduces the reporting frequency of low-priority nodes, decreasing the total amount of data transmitted over the network, effectively saving bandwidth resources and reducing the cloud data processing load. When environmental anomalies or drastic changes occur, the system can immediately increase the reporting priority of relevant nodes, ensuring that critical alarm information is transmitted preferentially and quickly, thereby significantly improving the real-time performance and reliability of the monitoring system.

[0016] 2. This invention utilizes time-division multiple access and dynamic time slot allocation technology in its communication resource scheduling module, combined with event-driven and timed wake-up node operating modes, to achieve orderly and conflict-free transmission scheduling of heterogeneous sensor data streams. This design not only avoids data packet contention and collisions during transmission, ensuring communication efficiency, but also allows sensor nodes to remain in a low-power sleep state most of the time, reducing the overall energy consumption of the monitoring system and facilitating long-term, stable, and unattended operation.

[0017] 3. This invention constructs a collaborative intelligent decision-making system from the edge to the cloud. Real-time, low-latency local decision-making and scheduling are performed at the edge, while the cloud utilizes stronger computing power for macro-level state prediction and strategy optimization. Through feedback from the state stability index, the system can adaptively adjust its monitoring sensitivity to better adapt to the operating characteristics of different racks and different periods. The strategy feedback optimization loop further enables the system to have self-evolution capabilities, continuously optimizing reporting strategies based on long-term operating data, thereby continuously improving the system's monitoring efficiency and adaptability, realizing a fundamental transformation of rack environment monitoring from static configuration to dynamic intelligent optimization. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall technical solution architecture of the IoT-based cabinet environment monitoring system proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the dynamic reporting strategy decision module in this invention; Figure 3 This is a schematic diagram of the multi-level interaction relationship and data flow of the communication resource scheduling module in this invention; Figure 4 This is a flowchart illustrating the collaborative working logic between the state assessment and prediction model and the edge computing gateway in this invention. Figure 5 This is a schematic diagram illustrating the principle framework of the event-driven and timed wake-up working mode of multiple types of sensor nodes in this invention; Detailed Implementation This embodiment details the technical implementation of an IoT-based cabinet environment monitoring system. Please refer to the appendix. Figure 1 The system consists of multiple types of sensor nodes deployed inside a rack, an edge computing gateway communicating with the sensor nodes, and a monitoring and analysis platform located in the cloud. The multiple types of sensor nodes include at least temperature and humidity sensors, vibration sensors, and smoke detectors. The edge computing gateway has built-in data acquisition and preprocessing modules, a dynamic reporting strategy decision-making module, and a communication resource scheduling module. The monitoring and analysis platform includes data receiving and storage units, a status assessment and prediction model, and a strategy parameter distribution interface. The following provides a detailed breakdown of the implementation of each module.

[0019] Multiple sensor nodes establish physical connections with the edge computing gateway via industrial-grade interfaces. The temperature and humidity sensors utilize digital integrated sensing chips, achieving a measurement accuracy of ±0.5℃ for temperature and ±3% relative humidity. The vibration sensor is based on a microelectromechanical system (MEMS) accelerometer with a range of -50G to 50G and a sampling frequency of 1000Hz. The smoke detector employs a photoelectric detection principle, achieving a detection sensitivity of 0.1% obscuration rate per meter. All sensor nodes are equipped with low-power microcontrollers and wireless communication modules, using the IEEE 802.15.4 standard. The sensor nodes are powered by 3.3V DC lithium batteries and support an energy harvesting module as an auxiliary power source.

[0020] The data acquisition and preprocessing module is deployed in the real-time operating system of the edge computing gateway. This module polls each sensor node via a multi-channel serial peripheral interface. It reads the raw data buffer of each sensor node every 2ms. The raw data includes temperature, humidity, triaxial vibration acceleration, and smoke concentration values. The preprocessing stage first applies a Kalman filter algorithm to the raw data to suppress high-frequency noise. The filtered data then enters a normalization unit, mapping it to a normalized range of 0 to 1. The normalization formula is: ; in For raw sensor data, and These represent the minimum and maximum values ​​of the sensor's range, respectively. The standardized data is encapsulated into fixed-length data frames, with the frame header containing the sensor node identifier, timestamp, and data checksum. The data frames are transmitted via the internal bus to the input queue of the dynamic reporting strategy decision module.

[0021] Please refer to the attached document. Figure 2The dynamic reporting strategy decision-making module consists of a threshold management unit, a change rate calculation unit, a multi-dimensional decision table unit, and an interval mapping unit. The threshold management unit presets a baseline state threshold range and an alarm state threshold for each type of sensor. Taking a temperature and humidity sensor as an example, the baseline state threshold range is 10℃ to 40℃ for temperature and 20% to 80% for humidity; the alarm state threshold is 50℃ for temperature and 90% for humidity. The change rate calculation unit calculates the absolute value of the difference rate between the current data and the previously reported data in real time. The calculation formula is: ; in This is the current sampled value. This is the previous valid reported value. The multidimensional decision table unit receives standardized data, the absolute value of the difference rate, and the threshold comparison results, and outputs the reporting priority level.

[0022] Priority is divided into three levels: Level 1 is the highest priority, triggered when data exceeds the baseline threshold and the absolute value of the difference rate is greater than 0.1; Level 2 is medium priority, triggered when data is within the baseline threshold but the absolute value of the difference rate is between 0.05 and 0.1; Level 3 is the lowest priority, triggered when data is stable and the absolute value of the difference rate is less than 0.05. The interval mapping unit queries the preset mapping relationship according to the priority level: Level 1 corresponds to a 2-second reporting interval, Level 2 corresponds to a 30-second reporting interval, and Level 3 corresponds to a 300-second reporting interval. When any sensor data exceeds the alarm status threshold, the priority is forcibly increased to Level 1, and an alarm data packet is generated.

[0023] Please refer to the attached document. Figure 3 The communication resource scheduling module employs a time-division multiple access (TDMA) mechanism. This module maintains a dynamic time slot allocation table, with each time slot having a length of 10ms. The time slot allocation table is divided into dedicated time slot areas and shared time slot areas. Level 1 priority nodes are allocated dedicated time slots, ensuring their data is transmitted to the gateway within 2 seconds. Level 2 and Level 3 priority nodes share the remaining time slots and coordinate transmission timing through a lightweight conflict avoidance algorithm.

[0024] The collision avoidance algorithm is based on carrier sense multiple access (CSMA) mechanism. Before transmission, nodes listen to the channel's busy / idle status. If the channel is idle, transmission occurs immediately; otherwise, a random backoff of 0 to 15 ms occurs. The communication resource scheduling module monitors alarm data packet markers in real time. Once a marker is detected, current non-urgent transmissions are immediately interrupted, and the highest-priority transmission channel is allocated to the alarm data. Data transmission uses forward error correction coding with a Reed-Solomon coding scheme, which can correct up to 8 bytes of errors.

[0025] Please refer to the attached document. Figure 4The state assessment and prediction model is deployed in the GPU cluster of the monitoring and analysis platform. This model is a deep learning model based on Long Short-Term Memory (LSTM) networks, with an input layer dimension of 128 and a hidden layer containing three LTM units, each with 256 neurons. The model training data consists of historical monitoring data from the past 30 days, with a sampling interval of 1 minute. The model outputs predicted values ​​and a state stability index for each sensor data point within the next 5 minutes. The state stability index is calculated based on the variance of the predicted sequence; the index is 0.9 when the variance is less than 0.01 and 0.3 when the variance is greater than 0.1. The monitoring and analysis platform periodically distributes the state stability index to the edge computing gateway via a policy parameter distribution interface.

[0026] After receiving the index, the dynamic reporting strategy decision module of the edge computing gateway adaptively adjusts the baseline state threshold range: when the index is higher than 0.8, the temperature threshold range is widened to 5℃ to 45℃ and the humidity threshold range is widened to 15% to 85%; when the index is less than 0.5, the temperature threshold range is narrowed to 15℃ to 35℃ and the humidity threshold range is narrowed to 30% to 70%.

[0027] Please refer to the attached document. Figure 5 The sensor nodes employ a hybrid event-driven and timed wake-up mode. During non-reporting periods, the nodes are in sleep mode with power consumption less than 10μA. Wake-up conditions include three scenarios: First, reaching the next reporting time point issued by the edge computing gateway, with a timing accuracy of ±1ms; second, the sensor locally detects a sudden change in a physical quantity, such as a temperature change rate greater than 0.5℃ per second or a vibration acceleration change greater than 10G; third, receiving an emergency polling command from the edge computing gateway, which is transmitted via a high-speed broadcast channel.

[0028] After being woken up, the node starts the data acquisition process. After the acquisition is completed, it sends the data to the gateway through the wireless module and then immediately returns to the sleep state.

[0029] The policy feedback optimization loop operates within the distributed computing framework of the monitoring and analysis platform. This loop collects operational data every 24 hours, including network bandwidth utilization, alarm response time, and average node energy consumption. The reinforcement learning algorithm is built upon a deep Q-network. The state space is a joint feature vector of bandwidth utilization, response time, and energy consumption, while the action space consists of threshold parameters and priority mappings in a multi-dimensional decision table. The reward function is designed as a linear combination of bandwidth utilization weighted at 0.3, response time weighted at 0.5, and energy consumption weighted at 0.2. The algorithm updates network parameters via gradient descent, and the optimized policy parameters are pushed to the edge computing gateway through a policy parameter distribution interface.

[0030] After system startup, the edge computing gateway initializes all sensor nodes and loads the default reporting policy. The data acquisition and preprocessing module continuously receives sensor data, and the dynamic reporting policy decision module performs priority calculations every 100ms. The communication resource scheduling module allocates time slots according to priority and sends scheduling instructions to the sensor nodes via wireless channels. The monitoring and analysis platform receives a batch of edge data every 5 minutes, and the state assessment and prediction model updates the prediction results in real time. The policy feedback optimization loop starts its optimization cycle at 0:00 every day, generating new policy parameters that take effect in the next cycle. Through the close collaboration of the above modules, the entire system achieves efficient, low-power, and intelligent adaptive monitoring of the rack environment.

[0031] This embodiment also provides an alternative implementation scheme for an IoT-based cabinet environment monitoring system, focusing on its adaptive mechanism under different network topologies and resource constraints. In this embodiment, water immersion sensors and door magnetic sensors are added to the multi-type sensor nodes to expand the monitoring dimensions. The water immersion sensor is based on the resistive detection principle and has a sensitivity of 0.1 mm water level change. The door magnetic sensor uses a Hall effect chip to detect the opening and closing status of the cabinet door. The edge computing gateway's hardware platform is upgraded to a multi-core processor, supporting parallel processing of 128 sensor data streams.

[0032] The data acquisition and preprocessing module introduces a wavelet transform filtering algorithm to replace the Kalman filter in Example 1. The wavelet basis function is db4, and the decomposition level is 4. The filtered data is then processed using the Z-score normalization method, with the following formula: ; in This represents the mean of the data within the sliding window. The standard deviation is used. The sliding window length is 100 sampling points. Standardized data are appended with a quality identifier, which includes a signal-to-noise ratio (SNR) and data integrity flags. Data with an SNR less than 10 dB is marked as low quality and used only for trend analysis without triggering a reporting strategy.

[0033] The dynamic reporting strategy decision module adds an environmental context awareness unit. This unit integrates rack location information and external meteorological data. For example, in hot weather, it automatically lowers the upper limit of the baseline state threshold of the temperature sensor by 5°C. The multi-dimensional decision table adds a time weighting factor, reducing the absolute value threshold for the difference rate of all sensors by 20% during peak business hours. The interval mapping unit supports non-linear mapping relationships; the reporting interval for the first-level priority can be dynamically adjusted between 1 second and 5 seconds, based on the network congestion index.

[0034] The network congestion index is calculated in real time by the communication resource scheduling module and is defined as the ratio of the number of currently active nodes to the total number of nodes.

[0035] The communication resource scheduling module supports dual-band transmission. The low-frequency band (868MHz) is used for Level 2 and Level 3 priority data transmission, while the high-frequency band (2.4GHz) is dedicated to Level 1 priority and alarm data. The time slot allocation table is expanded to a hierarchical structure, with a 5ms time slot length for the high-frequency band and a 20ms time slot length for the low-frequency band. The collision avoidance algorithm is upgraded to a hybrid time-division multiple access (TDMA) and frequency-division multiple access (FDMA) mechanism, with nodes automatically selecting the transmission frequency band based on priority. The gateway is equipped with a dual-antenna array, using beamforming technology to improve signal reception sensitivity.

[0036] The state assessment and prediction model integrates convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. CNNs extract spatial features (such as multi-sensor data correlations), while LTM networks capture temporal dependencies. An anomaly probability score is added to the model output; a score greater than 0.95 directly triggers a cloud-based manual review process. The policy parameter distribution interface supports differential encryption transmission to ensure the security of distributed parameters. The edge computing gateway adds a parameter verification unit; new policies can only be loaded after successful verification.

[0037] The sensor nodes employ an adaptive sampling rate mechanism. When the state stability index exceeds 0.9, the node's local sampling rate is reduced to 50% of its original value, resuming full-rate sampling only during reporting. The node sleep cycle is dynamically adjusted, with priority level 3 nodes having a sleep cycle that can be extended to 600 seconds. The gateway monitors the node's online status via a heartbeat mechanism, triggering an automatic reconnection protocol for lost nodes.

[0038] The policy feedback optimization loop employs a federated learning architecture. The edge computing gateway trains a lightweight policy model locally, uploading only the model gradients to the cloud. The cloud aggregates the gradients from each gateway, updates the global model, and then distributes the model parameters to each edge node. This mechanism reduces the amount of uploaded data by 90% and protects edge data privacy. The optimization cycle is shortened to 12 hours, improving the system's adaptability.

[0039] The edge computing gateway deploys a dual-machine hot backup system, with primary and backup nodes synchronizing status data via a high-speed bus. The communication link employs multi-path transmission technology, sending data packets through at least two independent paths. The cloud-based monitoring and analysis platform achieves multi-regional redundancy deployment, automatically switching to the backup center in the event of a single data center failure. Through these enhanced designs, the entire system maintains high robustness and scalability in complex environments.

[0040] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0041] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cabinet environment monitoring system based on the Internet of Things, characterized in that, This includes various types of sensor nodes deployed inside the cabinet, edge computing gateways that communicate with the sensor nodes, and a monitoring and analysis platform located in the cloud. Multiple types of sensor nodes include at least temperature and humidity sensors, vibration sensors, and smoke detectors; The edge computing gateway has a built-in data acquisition and preprocessing module, a dynamic reporting strategy decision-making module, and a communication resource scheduling module. The monitoring and analysis platform includes a data receiving and storage unit, a status assessment and prediction model, and a strategy parameter distribution interface; The data acquisition and preprocessing module is used to receive raw monitoring data from multiple types of sensor nodes and to filter and standardize the data. The dynamic reporting strategy decision module is used to calculate the data reporting priority and reporting time interval of each sensor node in real time based on the preprocessed data stream and the preset state evaluation rules. The communication resource scheduling module is used to allocate transmission time slots and control the data transmission behavior of sensor nodes based on the priority sequence and time interval parameters output by the dynamic reporting strategy decision module.

2. The rack environment monitoring system based on the Internet of Things according to claim 1, characterized in that, The operating mechanism of the dynamic reporting strategy decision module includes: setting a baseline state threshold range and an alarm state threshold for each type of sensor data; Calculate the absolute value of the rate of change of the current data relative to the previously reported data in real time; By combining whether the current data exceeds the baseline state threshold range and the magnitude of the absolute value of the rate of change, the reporting priority level of the sensor node is determined through a multi-dimensional decision table; The reporting priority level is divided into 3 levels, with level 1 being the highest priority and level 3 being the lowest priority; Based on the determined reporting priority level, query the preset priority and reporting time interval mapping relationship to obtain the specific time interval for the next data reporting.

3. The rack environment monitoring system based on the Internet of Things according to claim 2, characterized in that, When any sensor data exceeds its alarm status threshold, the dynamic reporting strategy decision module immediately forces the reporting priority of that node to level 1 and generates the highest priority alarm data packet. When the communication resource scheduling module detects the highest priority alarm data packet, it immediately interrupts the current non-urgent data transmission and prioritizes allocating a transmission channel for the alarm data packet.

4. The rack environment monitoring system based on the Internet of Things according to claim 1, characterized in that, The communication resource scheduling module uses a time-division multiple access mechanism to schedule data transmission; the module maintains a dynamic time slot allocation table; each sensor node is woken up and sends data in a specific time slot according to its assigned reporting time interval and priority; for nodes with priority level 1, the system reserves a dedicated time slot to ensure low-latency transmission; nodes with priority levels 2 and 3 share the remaining time slots and coordinate the transmission timing through a lightweight conflict avoidance algorithm.

5. The rack environment monitoring system based on the Internet of Things according to claim 1, characterized in that, The state assessment and prediction model runs on the monitoring and analysis platform. This model is a deep learning model based on long short-term memory networks. The model uses historical monitoring data to predict the future state of the cabinet environment; The model's output includes predicted values ​​of each sensor's data for the future, as well as a state stability index. The monitoring and analysis platform sends the state stability index to the edge computing gateway through the policy parameter distribution interface; The dynamic reporting strategy decision module of the edge computing gateway receives the state stability index and adaptively adjusts the baseline state threshold range accordingly.

6. The rack environment monitoring system based on the Internet of Things according to claim 1, characterized in that, The multi-type sensor nodes adopt a working mode that combines event-driven and timed wake-up; During non-reporting periods, sensor nodes are in a low-power sleep state; A sensor node is only awakened and performs data acquisition and reporting operations when one of the following three conditions is met: Condition 1, the next reporting time point specified by the edge computing gateway is reached; Condition 2: The physical quantity detected by the sensor itself undergoes a sudden change, and the magnitude of the change is greater than the preset local sudden change threshold. Condition 3: An emergency polling command is received from the edge computing gateway.

7. The rack environment monitoring system based on the Internet of Things according to claim 1, characterized in that, The monitoring and analysis platform is also equipped with a strategy feedback optimization loop; This loop continuously collects data on network bandwidth utilization, alarm response time, and node energy consumption during actual operation. Based on this operational data, the platform uses reinforcement learning algorithms to periodically optimize the multidimensional decision table and the mapping relationship between priority and reporting time interval in the dynamic reporting strategy decision module.

8. A cabinet environment monitoring system based on the Internet of Things according to claim 2, characterized in that, The calculation process of the absolute value of the rate of change is as follows: calculate the absolute value of the difference rate between the current data and the previous reported data in real time. The calculation formula is the absolute value of the current sampled value minus the previous valid reported value divided by the previous valid reported value.

9. A cabinet environment monitoring system based on the Internet of Things according to claim 4, characterized in that, The collision avoidance algorithm is based on the carrier sense multiple access mechanism, where nodes listen to the channel's busy / idle status before transmitting, and transmit immediately if the channel is idle.

10. A method for monitoring cabinet environment based on the Internet of Things, characterized in that, The cabinet environment monitoring system based on the Internet of Things as described in any one of claims 1 to 9 is used to realize cabinet environment monitoring.

Images