Debris flow monitoring and early warning system based on Internet of Things

Abstract

The invention relates to the technical field of geological disaster monitoring, and discloses a debris flow monitoring and early warning system based on Internet of Things, comprising a multi-source data acquisition module used for acquiring environmental data related to debris flow; the data transmission module is used for transmitting the environment data to a data center or a cloud end from a collection point; the data preprocessing module is used for carrying out abnormal value cleaning, time synchronization and normalization processing on the environment data; the deep learning analysis module is used for analyzing the preprocessed data so as to predict the occurrence risk of debris flow; and the dynamic early warning and publishing module is used for generating a multi-stage early warning signal according to an output result of the deep learning analysis module and publishing early warning information to a user in multiple modes. The accuracy, the real-time performance and the adaptability of the debris flow early warning system are improved through multi-source data fusion, edge calculation and a dynamic learning mechanism so as to cope with debris flow disasters in a complex geological environment.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring, and in particular to a debris flow monitoring and early warning system based on the Internet of Things. Background Art

[0002] As a typical geological disaster, debris flow is characterized by strong suddenness and great destructive power, posing a serious threat to residents and infrastructure in mountainous and hilly areas. In order to deal with the hazards brought by debris flow, the existing technology mainly adopts a monitoring method based on a single data source, which achieves debris flow early warning by installing rainfall sensors or seismic sensors and relying on a simple threshold trigger mechanism. However, this single data source monitoring method can only capture some of the triggering factors of debris flow, and cannot fully reflect the complex debris flow occurrence mechanism, resulting in low accuracy and timeliness of its prediction.

[0003] The existing warning system is constructed as follows: a single type of sensor is deployed in high-risk mountain areas, data is transmitted to the data center via wireless networks, and a simple statistical analysis method is used for warning. This system has many shortcomings: first, the monitoring method of a single data source cannot capture the complexity of debris flow triggering factors, which can easily lead to false alarms or missed alarms; second, the existing system lacks edge computing support and is difficult to achieve rapid response, and there is a delay in information processing; finally, the existing system lacks a continuous learning mechanism, and it is difficult to update the model based on real-time data, making it difficult for the warning system to adapt to the dynamic changes of the environment and climate. Therefore, how to use multi-source data fusion technology, real-time edge computing and continuous learning mechanisms to improve the accuracy, real-time and adaptability of debris flow warning systems has become an urgent problem to be solved in this field. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a debris flow monitoring and early warning system based on the Internet of Things, which solves the problems in the existing technology of low warning accuracy, poor real-time performance and insufficient adaptability to environmental changes caused by single data source monitoring, response delay and lack of dynamic adaptability.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a debris flow monitoring and early warning system based on the Internet of Things, comprising:

[0006] Multi-source data acquisition module, used to collect environmental data related to debris flow;

[0007] A data transmission module, used to transmit the environmental data from a collection point to a data center or a cloud;

[0008] A data preprocessing module, used for performing outlier cleaning, time synchronization and normalization processing on the environmental data;

[0009] A deep learning analysis module is used to analyze the preprocessed data to predict the risk of debris flow;

[0010] A dynamic warning and release module is used to generate multi-level warning signals according to the output results of the deep learning analysis module, and release warning information to users in various ways;

[0011] The edge computing and real-time response module is used to perform preliminary data analysis near the data collection point to provide quick response early warning prompts.

[0012] Preferably, the multi-source data acquisition module includes:

[0013] A rainfall sensor for collecting rainfall data;

[0014] Soil moisture sensor, used to collect soil moisture data;

[0015] Seismic sensors, used to collect information about ground vibrations;

[0016] Tilt sensor, used to collect surface tilt angle data.

[0017] Preferably, the data transmission module adopts a low-power wide area network protocol for data transmission, and the protocol includes a LoRa protocol or a NB-I oT protocol to ensure stable transmission of the environmental data in remote areas.

[0018] Preferably, the data preprocessing module includes:

[0019] An outlier detection unit, used to identify and delete abnormal data;

[0020] Missing value filling unit, used to interpolate and fill missing data;

[0021] Normalization unit is used to standardize data from different data sources to the same numerical range.

[0022] Preferably, the deep learning analysis module includes a spatiotemporal feature extraction layer and a fusion layer, wherein:

[0023] The spatiotemporal feature extraction layer is composed of a convolutional neural network module and a long short-term memory network module. The CNN module is used to extract the spatial features of the data, and the LSTM module is used to extract the time series features of the data.

[0024] The fusion layer is used to perform data fusion on the extracted spatiotemporal features to generate a feature vector for debris flow risk prediction.

[0025] Preferably, the deep learning analysis module further includes a multi-model integration unit for fusing the LSTM+CNN combination model with the integrated learning model, wherein the integrated learning model includes an XGBoost model and a random forest model to improve the prediction accuracy and robustness of the system.

[0026] Preferably, the dynamic warning and release module includes:

[0027] Dynamic threshold adjustment unit, used to automatically adjust the warning threshold according to rainfall, humidity increment, and surface inclination factors;

[0028] A multi-level warning unit, which is used to generate four-level warning signals: blue, yellow, orange and red, according to the risk level of debris flow;

[0029] The information release unit is used to release the warning signal to users through text messages, APP notifications, and emergency broadcasts.

[0030] Preferably, the dynamic threshold adjustment unit in the dynamic warning and release module uses the following formula to set the real-time warning threshold:

[0031] θ(t)=θ0+α·ΔR(t)+β·ΔS(t)+γ·ΔT(t)

[0032] Among them, θ0 is the basic threshold, α, β, γ are dynamically adjusted weight parameters, ΔR(t) is the rainfall increment, ΔS(t) is the soil moisture increment, and ΔT(t) is the change in surface inclination.

[0033] Preferably, the edge computing and real-time response module includes:

[0034] Edge computing units, which are used to perform preliminary data processing at nodes close to sensors to reduce data transmission delays;

[0035] Local prediction unit, which runs a simplified deep learning model on edge nodes to analyze environmental data in real time and achieve rapid early warning;

[0036] The data stream processing unit is used to batch process the high-frequency data stream of sensors to improve the real-time performance and response speed of data transmission.

[0037] Preferably, the system also includes a model updating unit for regularly updating the deep learning model through a continuous learning mechanism, wherein the updating method includes incremental training and full data retraining to adapt to environmental changes and improve prediction accuracy.

[0038] The present invention provides a debris flow monitoring and early warning system based on the Internet of Things. It has the following beneficial effects:

[0039] 1. The present invention integrates multi-dimensional environmental data such as rainfall, soil moisture, seismic activity, and surface inclination through a multi-source data acquisition module, and uses a deep learning analysis module to extract and fuse the spatiotemporal features of the data, comprehensively capturing the complex correlations among debris flow triggering factors. Through multi-source data fusion, the system can more accurately predict the probability of debris flow occurrence, greatly reduce false alarms and missed alarms caused by the limitations of a single data source, and significantly improve the accuracy of the prediction.

[0040] 2. The present invention deploys edge computing devices near sensor nodes and combines the multi-level warning mechanism of dynamic warning and release modules to enable the system to quickly respond to warnings in high-risk situations. Edge computing realizes the preliminary processing and rapid analysis of data, reduces transmission delays, and dynamic threshold adjustment automatically optimizes the warning threshold according to real-time environmental data, making the warning more flexible and accurate. This not only improves the response speed of the system, but also ensures the timeliness of the warning information, providing users with timely risk avoidance suggestions.

[0041] 3. The present invention uses a model update and continuous learning mechanism to enable the deep learning model to be continuously optimized based on newly added data, and regularly conduct incremental training and full retraining to ensure that the model can quickly adapt to environmental changes. Through incremental learning, the model can dynamically adjust parameters and maintain sensitivity to new environments and climate changes, thereby enhancing the long-term stability and robustness of the prediction model, which enables the system to still have efficient debris flow risk warning capabilities under different environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A perspective view of the present invention;

[0043] Figure 2 It is a structural schematic diagram of a multi-source data acquisition module of the present invention;

[0044] Figure 3 It is a structural schematic diagram of a data transmission module of the present invention;

[0045] Figure 4 It is a structural schematic diagram of a data preprocessing module of the present invention;

[0046] Figure 5 It is a structural schematic diagram of the deep learning analysis module of the present invention;

[0047] Figure 6 It is a structural schematic diagram of the dynamic warning and release module of the present invention;

[0048] Figure 7 It is a schematic diagram of the structure of the edge computing and real-time response module of the present invention;

[0049] Figure 8 It is a structural diagram of the continuous learning and model updating module of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example:

[0052] Please refer to the attached Figure 1 - Attachment Figure 8 The embodiment of the present invention provides a debris flow monitoring and early warning system based on the Internet of Things, which realizes real-time monitoring and early warning of debris flow disasters through Internet of Things technology, deep learning model, edge computing and multi-source data fusion. The system can realize comprehensive monitoring of the environment in high-risk areas, release disaster risk information to users through an intelligent early warning mechanism, and provide dynamic response.

[0053] 1. System overall architecture

[0054] The debris flow monitoring and early warning system consists of the following six modules: multi-source data acquisition module, data transmission module, data preprocessing module, deep learning analysis module, dynamic early warning and release module, edge computing and real-time response module. The functions and connection relationships of each module of the system are as follows:

[0055] Multi-source data acquisition module: responsible for collecting a variety of environmental data, such as rainfall, soil moisture, seismic activity, surface inclination, etc.;

[0056] Data transmission module: transmits the collected environmental data from the sensor node to the data center using a low-power network protocol;

[0057] Data preprocessing module: performs outlier cleaning, time synchronization and normalization on the collected data to provide high-quality input for the deep learning model;

[0058] Deep learning analysis module: Use LSTM and CNN models to extract spatiotemporal features of multi-source data and generate probability predictions for debris flow occurrence;

[0059] Dynamic warning and release module: According to the debris flow risk level predicted by the model, the warning threshold is adjusted in real time and warning information is released through multiple channels;

[0060] Edge computing and real-time response module: Perform preliminary data processing at nodes near the sensor to achieve rapid local early warning response.

[0061] 2. Multi-source data acquisition module

[0062] The module is deployed in areas with high risk of debris flow, and collects data from multiple sources in real time through multiple types of sensors. The main data sources include rainfall, soil moisture, seismic activity and surface inclination.

[0063] Rainfall sensor: Rain gauges are deployed in the target area to obtain rainfall and cumulative rainfall per unit time. Rainfall is one of the important triggering factors of debris flow. When the cumulative rainfall reaches a certain threshold, the saturation of the soil increases, and the risk of landslide increases accordingly.

[0064] Soil moisture sensor: This sensor is used to monitor soil moisture content. As rainfall continues, soil moisture increases. When the moisture reaches a critical value, the binding force between soil particles weakens, increasing the possibility of mudslides.

[0065] Seismic sensors: used to detect the intensity and frequency of ground vibrations. Earthquake activity can induce instability in the mountains. Seismic sensors can monitor vibration data and determine the impact of vibrations on mountain structures.

[0066] Inclination sensor: monitors the changes in the tilt angle of the ground in real time. The change in the tilt angle of a mountain often indicates the sliding of the mountain, especially a large increase in the tilt angle may lead to landslides and mudslides.

[0067] After the sensors are arranged, the data collected by each sensor will be timestamped to form a standardized data stream and uploaded to the data transmission module.

[0068] 3. Data transmission module

[0069] The data transmission module is responsible for transmitting multi-source sensor data from the collection point to the data center. Mountainous areas have complex terrain and limited power resources. Data transmission uses low-power, wide-area network protocols (such as LoRa or NB-IoT) to ensure stable data transmission.

[0070] Communication protocol: This system uses low-power wide area network protocols such as LoRa and NB-I oT to ensure stable data transmission in remote mountainous areas. LoRa has the characteristics of long distance and low power consumption, which is suitable for sensor nodes in mountainous areas. The high penetration of NB-I oT protocol is also suitable for low signal environments in mountainous areas.

[0071] Redundancy mechanism: The data transmission module is designed with a backup channel (such as satellite communication) as an alternative to cope with the interruption of the main transmission network and ensure the continuity of data transmission in harsh environments.

[0072] After the data is transmitted to the data center through the data transmission module, it enters the data preprocessing module for subsequent processing.

[0073] 4. Data preprocessing module

[0074] After multi-source data enters the data center, it needs to be preprocessed to ensure data quality. The data preprocessing module includes outlier cleaning, time synchronization, missing value filling, and normalization.

[0075] Outlier cleaning: Use the 3σ method to detect outliers and remove noise data that exceeds the normal range to ensure the authenticity and validity of the data.

[0076] Missing value filling: For missing data, linear interpolation or the mean of the previous and next time periods is used to fill in the missing data to ensure data continuity.

[0077] Time synchronization: Multi-source data has different sampling frequencies. The system unifies the time dimensions of different data sources through timestamps and synchronizes low-frequency data to the standard timeline.

[0078] Data normalization: Through Min-Max normalization or Z-score standardization, data of different dimensions are converted to the same numerical range to facilitate the training of deep learning models.

[0079] The processed data enters the deep learning analysis module for subsequent predictive analysis.

[0080] 5. Deep learning analysis module

[0081] The deep learning analysis module adopts a two-layer model architecture, including CNN and LSTM networks, to extract spatiotemporal features and perform risk analysis on multi-source data.

[0082] 5.1 Spatiotemporal Feature Extraction

[0083] CNN module: Convolutional neural network (CNN) is suitable for processing spatial features of data, such as the distribution information of sensors in geographic space. CNN is used to extract spatial features of sensor data such as rainfall and humidity, and obtain spatial correlation of various data sources.

[0084] LSTM module: Long short-term memory network (LSTM) is used to process time series data to capture the temporal characteristics of sensor data. LSTM is particularly suitable for processing the trend of rainfall and humidity over time and analyzing the dependency of data in the time dimension.

[0085] 5.2 Multi-model Ensemble

[0086] The deep learning module further integrates models based on spatiotemporal feature extraction. The integrated models include:

[0087] LSTM+CNN combined model: Through the coordinated use of CNN and LSTM, spatial features and temporal features are integrated.

[0088] Ensemble learning model: Combine the LSTM+CNN model with the XGBoost and random forest models to enhance the robustness and prediction stability of the model.

[0089] This module outputs the probability prediction value of debris flow through the deep learning model generated by training, providing data support for dynamic early warning.

[0090] 6. Dynamic warning and release module

[0091] Based on the output results of the deep learning analysis module, the dynamic warning and release module will generate multi-level warning signals and release them to users through multiple channels.

[0092] 6.1 Dynamic Threshold Adjustment

[0093] The warning threshold is adjusted dynamically according to the current environmental conditions and the debris flow risk probability output by the model. The calculation formula for the warning threshold is:

[0094] θ(t)=θ0+α·ΔR(t)+β·ΔS(t)+γ·ΔT(t)

[0095] Among them, θ0 is the basic threshold, α, β, and γ are weight coefficients, which respectively adjust the influence of rainfall increment ΔR(t), soil moisture increment ΔS(t), and surface inclination change ΔT(t) on the threshold.

[0096] 6.2 Multi-level warning generation

[0097] The system divides warning signals into four levels: blue, yellow, orange, and red according to the predicted debris flow risk level:

[0098] Blue warning: low risk, just remind you to pay attention to weather conditions;

[0099] Yellow warning: medium risk, reminding users to pay attention to safety;

[0100] Orange alert: high risk, residents are advised to take precautionary measures;

[0101] Red alert: extremely high risk, immediate evacuation and emergency avoidance are recommended.

[0102] 6.3 Information Release

[0103] The system releases warning information to users through SMS, APP notifications, broadcasts, etc. Through multi-channel release, it ensures that warning information is quickly delivered to users in the affected area, improving the efficiency of warning response.

[0104] 7. Edge computing and real-time response module

[0105] Preliminary data processing is performed on edge devices near the sensor nodes to improve the response speed and real-time performance of the system.

[0106] Edge computing devices: Edge computing devices are deployed near sensor nodes to perform preliminary data processing and reduce the delay caused by data transmission.

[0107] Local prediction: A simplified deep learning model runs on edge nodes to analyze data in real time and trigger preliminary warning signals immediately when high-risk events occur.

[0108] Data stream processing: Use data stream processing frameworks such as Kafka and Spark Streaming to process the high-frequency data streams generated by sensors in real time and optimize data transmission and processing efficiency.

[0109] 8. Model update and continuous learning mechanism

[0110] The system includes model updating and continuous learning modules to improve the prediction accuracy and adaptability of deep learning models.

[0111] Incremental training: The system performs incremental training based on newly collected data, continuously updates the model's parameters, and improves the model's adaptability to new environments.

[0112] Full retraining: All historical data is used to retrain the model regularly to maintain the model's prediction accuracy and long-term stability.

[0113] Data labeling and feedback: Through data labeling of actual debris flow events, feedback information is provided to the model to optimize the model's prediction effect.

[0114] 9. System workflow

[0115] The specific workflow of the system is as follows:

[0116] Data collection: Multi-source sensors collect environmental data in real time, including rainfall, soil moisture, seismic data, and changes in ground surface tilt.

[0117] Data transmission: Data is transmitted to the data center via LoRa or NB-I oT protocol.

[0118] Data preprocessing: The data preprocessing module cleans, synchronizes and standardizes the raw data to ensure data quality.

[0119] Deep learning analysis: The deep learning analysis module extracts spatiotemporal features from preprocessed data to predict debris flow risks.

[0120] Dynamic warning: The dynamic warning module adjusts the warning threshold in real time according to the prediction results and generates the corresponding warning level.

[0121] Information release: Release warning information to users via SMS, APP and broadcast.

[0122] Emergency response: In the case of high-risk warning, the system automatically notifies the emergency management department to coordinate evacuation and rescue.

[0123] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A debris flow monitoring and early warning system based on the Internet of Things, characterized in that: include Multi-source data acquisition module, used to collect environmental data related to debris flow; A data transmission module, used to transmit the environmental data from a collection point to a data center or a cloud; A data preprocessing module, used for performing outlier cleaning, time synchronization and normalization processing on the environmental data; A deep learning analysis module is used to analyze the preprocessed data to predict the risk of debris flow; A dynamic warning and release module, which is used to generate multi-level warning signals according to the output results of the deep learning analysis module, and release warning information to users in various ways; The edge computing and real-time response module is used to perform preliminary data analysis near the data collection point to provide quick response early warning prompts.

2. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The multi-source data acquisition module comprises: A rainfall sensor for collecting rainfall data; Soil moisture sensor, used to collect soil moisture data; Seismic sensors, used to collect information about ground vibrations; Tilt sensor, used to collect surface tilt angle data.

3. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The data transmission module adopts a low-power wide area network protocol for data transmission, and the protocol includes a LoRa protocol or a NB-IoT protocol to ensure stable transmission of the environmental data in remote areas.

4. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The data preprocessing module comprises: An outlier detection unit, used to identify and delete abnormal data; Missing value filling unit, used to interpolate and fill missing data; Normalization unit is used to standardize data from different data sources to the same numerical range.

5. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The deep learning analysis module includes a spatiotemporal feature extraction layer and a fusion layer, wherein: The spatiotemporal feature extraction layer is composed of a convolutional neural network module and a long short-term memory network module. The CNN module is used to extract the spatial features of the data, and the LSTM module is used to extract the time series features of the data. The fusion layer is used to perform data fusion on the extracted spatiotemporal features to generate a feature vector for debris flow risk prediction.

6. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The deep learning analysis module further includes a multi-model integration unit for fusing the LSTM+CNN combination model with an integrated learning model, wherein the integrated learning model includes an XGBoost model and a random forest model to improve the prediction accuracy and robustness of the system.

7. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The dynamic warning and release module includes: Dynamic threshold adjustment unit, used to automatically adjust the warning threshold according to rainfall, humidity increment, and surface inclination factors; A multi-level warning unit, which is used to generate four-level warning signals: blue, yellow, orange and red, according to the risk level of debris flow; The information release unit is used to release the warning signal to users through text messages, APP notifications, and emergency broadcasts.

8. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The dynamic threshold adjustment unit in the dynamic warning and release module uses the following formula to set the real-time warning threshold: θ(t)=θ0+α·ΔR(t)+β·ΔS(t)+γ·ΔT(t) Among them, θ0 is the basic threshold, α, β, γ are dynamically adjusted weight parameters, ΔR(t) is the rainfall increment, ΔS(t) is the soil moisture increment, and ΔT(t) is the change in surface inclination.

9. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1 is characterized in that: The edge computing and real-time response module includes: Edge computing units, which are used to perform preliminary data processing at nodes close to sensors to reduce data transmission delays; Local prediction unit, which runs a simplified deep learning model on edge nodes to analyze environmental data in real time and achieve rapid early warning; The data stream processing unit is used to batch process the high-frequency data stream of sensors to improve the real-time performance and response speed of data transmission.

10. The debris flow monitoring and early warning system based on the Internet of Things according to claim 1, characterized in that: The system also includes a model updating unit, which is used to regularly update the deep learning model through a continuous learning mechanism. The updating method includes incremental training and full data retraining to adapt to environmental changes and improve prediction accuracy.

Images