Unmanned tower operation status monitoring system based on big data

Abstract

The present invention relates to the technical field of unmanned tower operation status monitoring technology and discloses a big data-based unmanned tower operation status monitoring system. The system includes a data acquisition unit that collects parameters such as the unmanned tower's communication link, weather, navigation, power, and aircraft trajectory in real time; a multi-source data fusion unit that uses a federated learning algorithm to perform data fusion and generate a multimodal joint feature matrix; a dynamic modeling unit that constructs a state transition probability graph based on a dynamic Bayesian network and outputs a real-time state evolution sequence; a real-time anomaly detection unit that calculates anomaly confidence levels using an isolation forest algorithm and an LSTM autoencoder; an active control instruction generation unit that generates a device control instruction set based on anomaly confidence levels combined with reinforcement learning; and a knowledge graph reasoning unit that optimizes the state transition probability graph. This system can comprehensively, in real time, and accurately monitor and control the operation status of the unmanned tower, improving operational safety and stability and reducing operational risks.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned tower operation status monitoring, and in particular to an unmanned tower operation status monitoring system based on big data. Background Art

[0002] With the rapid development of aviation, unmanned control towers (UCTs) have shown great potential in improving airport operational efficiency and reducing labor costs. However, the safe monitoring of their operational status faces numerous challenges. Traditional manned control towers rely on human operators, who can visually observe and operate various equipment. However, unmanned control towers lack real-time human intervention. Equipment failures or malfunctions cannot be detected and addressed promptly, potentially leading to serious safety incidents. For example, a communication link failure can disrupt information transmission between the control tower and aircraft, paralyzing command and dispatch. Sudden changes in weather conditions can threaten the safety of aircraft takeoffs and landings if accurate weather data cannot be obtained in a timely manner.

[0003] Existing technologies have significant shortcomings in data collection and processing. Data collected by different types of sensors and equipment uses varying formats and standards, creating data silos that hinder effective integration and analysis. For example, there's a lack of correlation analysis between communication link status parameters, meteorological sensor monitoring parameters, and navigation equipment operating parameters, making it impossible to comprehensively understand the operational status of the unmanned tower. Furthermore, faced with massive amounts of real-time data, traditional data processing methods are inefficient, unable to meet real-time requirements, and struggle to quickly and accurately identify potential anomalies.

[0004] Existing monitoring systems typically use simple threshold-based methods for anomaly detection and resolution, resulting in high false alarm and missed alarm rates. This method is unable to comprehensively determine the presence of an anomaly when multiple parameters undergo subtle changes simultaneously, leading to potential faults being overlooked. Furthermore, after an anomaly is detected, there is a lack of effective proactive control strategies, making it impossible to precisely adjust the equipment in a timely manner to properly resolve the issue. For example, when navigation equipment experiences a decrease in accuracy, effective calibration measures cannot be quickly developed and implemented, compromising the aircraft's navigation accuracy and flight safety.

[0005] Furthermore, existing unmanned tower monitoring systems lack in-depth understanding of the interdependencies between devices. Each device does not operate in isolation but rather is interconnected and mutually influential. However, existing technologies cannot fully leverage these interdependencies for comprehensive risk assessment and prediction, making it difficult to proactively identify potential failure risks and providing robust support for equipment maintenance and management. Therefore, there is an urgent need for an unmanned tower operating status monitoring system based on big data to address these issues and ensure the safe and stable operation of unmanned towers. Summary of the Invention

[0006] The purpose of the present invention is to provide an unmanned tower operation status monitoring system based on big data to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an unmanned tower operation status monitoring system based on big data, the system comprising:

[0008] A data acquisition unit, which is used to collect communication link status parameters of the unmanned tower, meteorological sensor monitoring parameters, navigation equipment operating parameters, power equipment load parameters, and aircraft dynamic trajectory parameters in real time through a stream data processing framework;

[0009] a multi-source data fusion unit, configured to perform cross-domain feature extraction and heterogeneous data fusion on the communication link status parameters, meteorological sensor monitoring parameters, navigation equipment operating parameters, power equipment load parameters, and aircraft dynamic trajectory parameters through a federated learning algorithm to generate a multi-modal joint feature matrix;

[0010] A dynamic modeling unit, configured to construct an unmanned tower operation state transition probability graph based on a dynamic Bayesian network and taking the multimodal joint feature matrix as input, and output a real-time state evolution sequence;

[0011] A real-time anomaly detection unit, configured to calculate a time series anomaly score for the real-time state evolution sequence using an isolation forest algorithm and an LSTM autoencoder to generate an anomaly confidence level;

[0012] An active control instruction generation unit is used to generate a device control instruction set based on the abnormality confidence and in combination with a reinforcement learning strategy network.

[0013] Preferably, the execution steps of the real-time anomaly detection unit include:

[0014] Dividing the real-time state evolution sequence into time window segments, and extracting statistical feature vectors within each time window;

[0015] Performing reconstruction error calculation on the statistical feature vector using the LSTM autoencoder to obtain a first anomaly score;

[0016] Performing path length deviation calculation on the statistical feature vector using the isolation forest algorithm to obtain a second anomaly score;

[0017] The first anomaly score and the second anomaly score are weightedly fused to generate the anomaly confidence score.

[0018] Preferably, the execution steps of the data acquisition unit include:

[0019] Receive signal strength, bit error rate, and link delay data from the communication link status parameters in real time via the Apache Kafka stream processing platform and the communication protocol interface of the unmanned tower;

[0020] Collecting wind speed, visibility and pressure gradient data among the meteorological sensor monitoring parameters through distributed sensor nodes;

[0021] The gyroscope drift value, GPS positioning deviation and radar scanning frequency data in the navigation device operating parameters are extracted through the device log parsing engine.

[0022] Preferably, the execution step of the multi-source data fusion unit further includes:

[0023] Build a global federated learning model, where each tower device acts as a local client and uploads encrypted local feature gradients;

[0024] Performing secure weighted averaging on the local feature gradients through an aggregation server to update the global model parameters;

[0025] The updated global model parameters are sent to each client, and the multimodal joint feature matrix is ​​iteratively generated.

[0026] Preferably, the execution steps of the dynamic modeling unit include:

[0027] Defining node variables of the dynamic Bayesian network based on historical operating data, including communication link stability, meteorological interference level, navigation precision reduction factor, power load fluctuation index, and aircraft trajectory offset;

[0028] Calculating the conditional probability distribution of the node variables by using a Markov chain Monte Carlo method;

[0029] The edge weights of the state transition probability graph are updated based on the real-time data stream to generate the real-time state evolution sequence.

[0030] Preferably, the execution step of the real-time anomaly detection unit further includes:

[0031] The hidden layer dimension of the LSTM autoencoder is set to 30% of the input feature dimension, and a Dropout layer is used to prevent overfitting;

[0032] An adaptive subtree splitting strategy is introduced into the isolation forest algorithm to dynamically adjust the abnormal segmentation threshold according to data distribution.

[0033] Preferably, the execution steps of the active control instruction generating unit include:

[0034] Constructing a state space of a reinforcement learning policy network, including the anomaly confidence level, the equipment health index, and the environmental risk level;

[0035] The reward function is defined as the entropy reduction of the state transfer matrix after device regulation.

[0036] The policy network parameters are iteratively updated through a proximal policy optimization algorithm, and the device control instruction set is output.

[0037] Preferably, the step of generating the device control instruction set further includes:

[0038] Generate emergency shutdown instructions for high-priority abnormal events and trigger redundant equipment switching protocols;

[0039] Generate dynamic parameter compensation instructions for medium and low priority abnormal events, including communication link power adjustment, navigation equipment calibration coefficient correction and power load balancing distribution.

[0040] Preferably, the system further comprises:

[0041] The knowledge graph reasoning unit is used to build an unmanned tower equipment association graph based on a graph neural network and extract key node influencing factors;

[0042] The execution steps of the knowledge graph reasoning unit include:

[0043] Mapping the communication link status parameters, navigation equipment operating parameters and power equipment load parameters into graph nodes;

[0044] Calculate the dependency strength between nodes through the graph attention mechanism to generate the key node influence factor;

[0045] The key node influencing factors are input into the dynamic modeling unit to optimize the state transition probability graph.

[0046] Preferably, the present invention further includes an electronic device, comprising:

[0047] A memory for storing a computer program for the unmanned tower operation status monitoring system based on big data;

[0048] The processor is used to execute the computer program to realize real-time monitoring and control of the operating status of the unmanned tower.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The unmanned tower operation status monitoring system based on big data of the present invention has many significant beneficial effects. In terms of data collection and fusion, the system uses a stream data processing framework and a federated learning algorithm to collect multi-source heterogeneous data in real time and perform efficient fusion. The data acquisition unit can obtain parameters such as communication links, meteorology, navigation, electricity, and aircraft trajectories in a variety of ways, breaking down data silos and allowing various types of data to work together. The multi-source data fusion unit uses a federated learning algorithm to complete cross-domain feature extraction and heterogeneous data fusion while protecting data privacy, and generates a multimodal joint feature matrix. This enables the system to have a more comprehensive and accurate understanding of the operating status of the unmanned tower, providing a solid data foundation for subsequent analysis and decision-making. For example, after fusing the communication link status parameters with the meteorological sensor monitoring parameters, the impact of meteorological conditions on communication quality can be analyzed, and countermeasures can be taken in advance to ensure the stability of communication.

[0051] The system boasts outstanding dynamic modeling and anomaly detection capabilities. The dynamic modeling unit, based on a dynamic Bayesian network, uses fused data as input to construct an operating state transition probability graph and outputs a real-time state evolution sequence, which accurately captures dynamic changes in the system's operating state. The real-time anomaly detection unit combines the isolation forest algorithm with an LSTM autoencoder to calculate anomaly scores for the state evolution sequence and generate anomaly confidence levels. This multi-algorithm collaborative approach greatly improves the accuracy and timeliness of anomaly detection. By dividing the sequence into time window segments to extract statistical feature vectors, and then using two algorithms to calculate anomaly scores and fuse them, it can effectively identify a variety of complex anomalies. For example, during navigation equipment operation, even subtle parameter changes can promptly detect potential anomalies, significantly reducing false alarm and missed detection rates compared to traditional threshold judgment methods.

[0052] The active control instruction generation mechanism is efficient and practical. The active control instruction generation unit generates a set of device control instructions based on the anomaly confidence level and the reinforcement learning policy network. A reasonable state space and reward function are constructed, and the policy network parameters are iteratively updated through the proximal policy optimization algorithm, so that the generated control instructions are accurate and effective. The system will take different measures for abnormal events of different priorities. When a high-priority abnormal event occurs, an emergency shutdown instruction is quickly generated and a redundant equipment switching protocol is triggered to ensure the continuous operation of key functions; medium and low-priority abnormal events generate parameter dynamic compensation instructions, such as adjusting the communication link power, correcting the navigation equipment calibration coefficient, balancing the power load distribution, etc., to promptly repair the abnormality without affecting the overall operation and avoid the deterioration of the problem.

[0053] The knowledge graph reasoning unit further optimizes system performance. Based on a graph neural network, this unit constructs a correlation graph of unmanned tower equipment, extracts influencing factors of key nodes, and then inputs this information into the dynamic modeling unit to optimize the state transition probability graph. This enables the system to deeply explore the relationships between equipment and conduct risk assessment and prediction from a global perspective. For example, by analyzing the relationship between communication link nodes and navigation equipment nodes, the risk of navigation anomalies caused by communication failures can be predicted in advance, allowing maintenance or preventive measures to be scheduled in advance, thereby improving system reliability and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a working principle diagram of the unmanned tower operation status monitoring system based on big data according to the present invention;

[0055] Figure 2 A workflow diagram for model completion and data sharing for generating adversarial networks;

[0056] Figure 3 A workflow diagram for the reward function and scene migration of the multi-objective reinforcement learning algorithm;

[0057] Figure 4 This is the workflow diagram of spatiotemporal alignment technology and abnormal data detection. DETAILED DESCRIPTION

[0058] 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.

[0059] See also Figure 1-Figure 4 The present invention provides an unmanned tower operation status monitoring system based on big data, which aims to achieve real-time, accurate monitoring and effective control of the unmanned tower operation status. The specific implementation scheme of the present invention is described in detail below.

[0060] The system primarily consists of a data acquisition unit, a multi-source data fusion unit, a dynamic modeling unit, a real-time anomaly detection unit, and an active control command generation unit. In actual operation, these units work together to ensure the stable operation of the unmanned tower.

[0061] The data acquisition unit uses a streaming data processing framework to collect real-time data from the unmanned tower's communication link status parameters, meteorological sensor monitoring parameters, navigation equipment operating parameters, power equipment load parameters, and aircraft dynamic trajectory parameters. For example, through specific sensors and data interfaces, it connects to various devices in the unmanned tower to continuously obtain these key parameters, providing basic data for subsequent analysis and processing.

[0062] The multi-source data fusion unit uses a federated learning algorithm to extract cross-domain features from various collected parameters and fuse heterogeneous data to generate a multimodal joint feature matrix. Federated learning algorithms can achieve data fusion across different data sources while protecting data privacy, thereby increasing the value of data utilization.

[0063] The dynamic modeling unit, based on a dynamic Bayesian network, uses a multimodal joint feature matrix as input to construct a probability graph of the unmanned tower's operational state transitions and output a real-time state evolution sequence. The dynamic Bayesian network effectively captures dynamic changes in system states, providing a basis for subsequent anomaly detection.

[0064] The real-time anomaly detection unit uses the Isolation Forest algorithm and LSTM autoencoder to calculate a time series anomaly score for the real-time state evolution sequence and generate anomaly confidence. The combination of these two algorithms can more accurately identify anomalies in the system, improving the accuracy and reliability of anomaly detection.

[0065] The active control command generation unit generates a set of equipment control commands based on the anomaly confidence level and a reinforcement learning strategy network. When an anomaly is detected, the unit quickly generates corresponding control commands to adjust the unmanned tower equipment to ensure normal operation of the system.

[0066] Through the coordinated work of the above-mentioned units, the present invention can achieve comprehensive, real-time monitoring and precise control of the operating status of the unmanned tower, thereby improving the operational safety and stability of the unmanned tower. Example 1

[0067] The real-time anomaly detection unit is responsible for real-time monitoring of the operating status of the unmanned tower and timely detection of potential anomalies.

[0068] First, the real-time state evolution sequence is divided into time window segments, and the statistical feature vectors in each time window are extracted. Assuming that the time window length is , in this time window, the real-time state evolution sequence data collected is Statistical eigenvectors include common statistics such as mean and variance. The mean calculation formula is: in, represents the mean in the statistical eigenvector, Indicates the time window Status data at a moment, is the length of the time window. The variance calculation formula is: in, Represents the variance in the statistical feature vector. The reconstruction error of the statistical feature vector is calculated by the LSTM autoencoder to obtain the first anomaly score. The LSTM autoencoder can learn the feature representation of the statistical feature vector under normal conditions. When the input statistical feature vector deviates from the normal state, the reconstruction error will increase. Let the reconstruction error be , The larger the value is, the higher the first anomaly score is, that is, the greater the possibility of an anomaly.

[0069] The path length deviation of the statistical feature vector is calculated using the Isolation Forest algorithm to obtain the second anomaly score. The Isolation Forest algorithm determines the degree of anomaly of a data point by constructing a tree structure.

[0070] The first anomaly score and the second anomaly score are weighted and fused to generate anomaly confidence. Assume that the weight of the first anomaly score is , the weight of the second anomaly score is ,and , the calculation formula of abnormal confidence is: in, represents the abnormal confidence, is the first anomaly score, Score the second anomaly, and is the weight coefficient.

[0071] In terms of data collection, the data acquisition unit uses the Apache Kafka stream processing platform to communicate with the unmanned tower via a communication protocol interface, receiving real-time data on communication link status parameters such as signal strength, bit error rate, and link latency. Distributed sensor nodes also collect wind speed, visibility, and pressure gradient data from meteorological sensors. A device log parsing engine extracts gyroscope drift, GPS positioning deviation, and radar scanning frequency data from navigation equipment operating parameters. This collected data provides fundamental data support for the real-time anomaly detection unit, ensuring the accuracy and reliability of anomaly detection. Example 2

[0072] The role of the multi-source data fusion unit is to fuse data from different data sources with different formats and features to generate a more valuable multi-modal joint feature matrix, providing high-quality data support for subsequent dynamic modeling and anomaly detection. Specific methods include:

[0073] Construct a global federated learning model. In this model, each tower device acts as a local client and uploads encrypted local feature gradients. Assume that there is local clients, each client Have local data , local data is trained locally using a machine learning algorithm (such as a neural network) to obtain a local model , and calculate the local feature gradient Here It is the gradient vector about the local model parameters, which reflects the direction and degree of influence of local data on the model parameters.

[0074] The aggregation server performs a secure weighted average of local feature gradients to update global model parameters. The aggregation server will assign a weighted average to each local feature gradient based on factors such as the importance or data volume of each client. Assign a weight , the weights satisfy and The update formula of the global model parameters is:

[0075] in, Represents the updated global feature gradient, which integrates the feature gradient information of each local client. Based on this updated global feature gradient, the aggregation server updates the parameters of the global model to obtain a new global model .

[0076] The updated global model parameters are sent to each client to iteratively generate a multimodal joint feature matrix. After receiving the updated global model parameters, each client uses local data to further train and optimize the model. In this process, the model will continuously learn the associations and features between data from different data sources, and gradually generate a multimodal joint feature matrix. Assume that after The multimodal joint feature matrix generated after iteration is In each iteration, the client will adjust the The federated learning process is updated to more accurately reflect the fusion characteristics of different data. As the number of iterations increases, the quality of the multimodal joint feature matrix continues to improve, providing a more accurate data foundation for subsequent dynamic modeling and anomaly detection. This federated learning process ensures data privacy while achieving efficient fusion of multi-source data, enhancing the system's ability to monitor the operating status of unmanned towers. Example 3

[0077] The dynamic modeling unit builds a dynamic Bayesian network to model the operating status of the unmanned tower and generate a real-time state evolution sequence, providing a key basis for real-time anomaly detection and active control command generation. Specific methods include:

[0078] Based on historical operational data, the node variables of the dynamic Bayesian network are defined, including communication link stability, weather interference level, navigation DOP, power load fluctuation index, and aircraft trajectory deviation. Communication link stability measures the reliability of the communication link during data transmission; weather interference level reflects the impact of weather conditions on the operation of the unmanned tower; navigation DOP reflects the degradation of navigation equipment accuracy over time; power load fluctuation index describes the changes in power equipment load; and aircraft trajectory deviation indicates the degree to which the actual aircraft trajectory deviates from the planned trajectory. These node variables comprehensively reflect the key characteristics of the unmanned tower's operational status.

[0079] The conditional probability distribution of node variables is calculated by Markov chain Monte Carlo method. Markov chain Monte Carlo method is a random sampling algorithm based on Markov chain, which is used to approximate probability distribution in high-dimensional space. Assume that there are nodes in the dynamic Bayesian network. and ,node The value of ,node The value of , then the node At a given node Conditional probability distribution under conditions It can be estimated by the Markov chain Monte Carlo method. Specifically, by constructing a Markov chain, its stationary distribution is the target conditional probability distribution , and then sample from this Markov chain to obtain a series of samples, and calculate based on these samples For example, when calculating the conditional probability distribution of communication link stability under the influence of meteorological interference levels, we use historical data to construct a Markov chain. After multiple sampling and statistics, we can obtain the probability of the communication link being in different stable states under different meteorological interference levels.

[0080] Based on the real-time data stream, the edge weights of the state transition probability graph are updated to generate a real-time state evolution sequence. The real-time data stream is continuously input into the dynamic modeling unit, and the edge weights between nodes in the dynamic Bayesian network are updated according to the new data. The edge weight reflects the strength of the dependency between nodes. The larger the weight, the closer the relationship between the two nodes. Suppose that the node in the state transition probability graph is and nodes The edge weight between , when there is new real-time data indicating that the node and nodes When the dependency between As edge weights are updated, the dynamic Bayesian network continuously evolves, generating a real-time state evolution sequence that reflects the real-time operating status of the unmanned tower. This real-time state evolution sequence provides an accurate reference for subsequent real-time anomaly detection, helping the system promptly identify potential anomalies. Through these steps, the dynamic modeling unit can effectively dynamically model and analyze the operating status of the unmanned tower, providing strong support for the stable operation of the entire monitoring system. Example 4

[0081] The real-time anomaly detection unit is used to ensure the safe and stable operation of the unmanned tower, and the active control instruction generation unit is the execution unit that takes effective measures when an abnormal situation occurs. The two work closely together to ensure the reliability of the system.

[0082] In terms of real-time anomaly detection units, the hidden layer dimension of the LSTM autoencoder is set to 30% of the input feature dimension, and the Dropout layer is used to prevent overfitting. When the LSTM autoencoder reconstructs the statistical feature vector, the hidden layer plays a key role in feature extraction and transformation. Assuming the input feature dimension is , then the hidden layer dimension is set to This setting was determined after a lot of experiments and data analysis. It can effectively extract features and avoid overfitting problems caused by overly complex models. The Dropout layer randomly discards some neurons during the training process, so that the model will not be overly dependent on certain specific neuron connections, thereby improving the generalization ability of the model. Let the dropout layer's dropout probability be , usually takes the value Between, for example During the training process, for each neuron in the LSTM autoencoder, the probability Setting its output to 0 can reduce the co-adaptation phenomenon between neurons and enhance the robustness of the model.

[0083] The Isolation Forest algorithm introduces an adaptive subtree splitting strategy to dynamically adjust the anomaly segmentation threshold based on data distribution. Traditional Isolation Forest algorithms use a fixed splitting strategy and threshold when constructing the tree structure, which may not be well adapted to complex data distributions. However, the adaptive subtree splitting strategy dynamically selects the optimal split point and adjusts the anomaly segmentation threshold based on the distribution of data points. Assuming that at a certain moment, the data distribution in a dataset exhibits a specific pattern, the algorithm analyzes the data, calculates the splitting gain of each potential split point, and selects the point with the largest splitting gain as the splitting point. Furthermore, the anomaly segmentation threshold is dynamically adjusted based on the data's degree of dispersion and distribution range. This adaptive strategy enables the Isolation Forest algorithm to more accurately identify anomalous data points, improving the accuracy of anomaly detection.

[0084] In terms of the active control instruction generation unit, the state space of the reinforcement learning strategy network is constructed, including anomaly confidence, equipment health index and environmental risk level. Anomaly confidence is an indicator output by the real-time anomaly detection unit that reflects the possibility of abnormal situations; the equipment health index is used to measure the operating health of each device in the unmanned tower, and is obtained by a comprehensive evaluation of the various performance parameters of the equipment; the environmental risk level is a classification of potential risks to the operation of the unmanned tower based on factors such as meteorological conditions and the surrounding electromagnetic environment. The reward function is defined as the entropy reduction amplitude of the state transfer matrix after the equipment is controlled. Entropy is an indicator of system uncertainty. The greater the entropy reduction amplitude of the state transfer matrix, the more the uncertainty of the system is reduced after the equipment is controlled, and the better the control effect. Let the state transfer matrix be , whose entropy is , the reward function The calculation formula is: in, represents the entropy of the state transfer matrix before device regulation, Represents the entropy of the state transition matrix after device control. The proximal policy optimization algorithm iteratively updates the policy network parameters and outputs a set of device control instructions. While ensuring stable policy updates, the proximal policy optimization algorithm rapidly optimizes the policy network, enabling it to generate optimal device control instructions based on varying state space information. During this process, the policy network parameters are continuously adjusted based on feedback from the reward function to improve device control effectiveness and ensure stable operation of the unmanned tower. Example 5

[0085] This example primarily describes the execution steps of the knowledge graph reasoning unit and the detailed generation of the equipment control instruction set. The knowledge graph reasoning unit provides a knowledge-association-based analysis method for the unmanned tower operation status monitoring system, while the generation of the equipment control instruction set is directly related to the effective control of unmanned tower equipment in abnormal situations.

[0086] The knowledge graph inference unit is used to construct an unmanned tower equipment association graph based on a graph neural network and extract the influencing factors of key nodes. First, communication link status parameters, navigation equipment operating parameters, and power equipment load parameters are mapped into graph nodes. For example, parameters such as the communication link's signal strength and bit error rate are used as attributes of the communication link-related nodes; parameters such as the navigation equipment's gyroscope drift value and GPS positioning deviation are used as attributes of the navigation equipment node; and parameters such as the power equipment's load fluctuation index are used as attributes of the power equipment node. These nodes are interconnected through physical or logical connections to form a device association graph.

[0087] Next, the graph attention mechanism calculates the inter-node dependency strength and generates key node influence factors. The graph attention mechanism automatically learns the importance weights between nodes, highlighting nodes with the greatest impact on the target node. By calculating the dependency strengths between all nodes, the key node influence factors are derived. These key node influence factors indicate which nodes have the greatest influence within the entire device association graph, providing important insights for subsequent system analysis and optimization.

[0088] The dynamic modeling unit inputs the key node influencing factors into the state transition probability diagram to optimize it. This consideration of key node influencing factors when constructing the state transition probability diagram allows the dynamic modeling unit to more accurately reflect the transition relationships between the unmanned tower's operating states. For example, if a communication link node is identified as a key node, and its state changes significantly impact the overall system state transition, the edge weights associated with that node in the state transition probability diagram will be adjusted accordingly based on its influencing factor, making the state transition probability diagram more realistic.

[0089] In terms of generating equipment control instruction sets, emergency shutdown instructions are generated for high-priority abnormal events, and redundant equipment switching protocols are triggered. When the abnormality confidence exceeds the preset high-priority threshold and the system comprehensively judges it to be a high-priority abnormal event, an emergency shutdown instruction is immediately generated to stop the operation of related equipment to prevent further expansion of the fault. At the same time, the redundant equipment switching protocol is triggered to start the backup equipment to ensure that the key functions of the unmanned tower can continue to operate. For example, when a serious failure of the navigation equipment may cause a serious decline in the navigation accuracy of the aircraft and endanger flight safety, the system will quickly issue an emergency shutdown instruction, stop the faulty navigation equipment, and automatically switch to the backup navigation equipment to ensure the continuity of the navigation function.

[0090] Generate dynamic parameter compensation instructions for medium and low priority abnormal events, including communication link power adjustment, navigation equipment calibration coefficient correction and power load balancing distribution. When medium and low priority abnormal events occur, the impact of the abnormality is compensated by adjusting the parameters of related equipment. In terms of communication links, if the signal strength drops, according to the degree of drop and communication requirements, according to the formula Adjust the communication link power, where is the adjusted power, is the power before adjustment, It is the power adjustment calculated based on the abnormal situation. In this way, the signal strength is enhanced to ensure the communication quality. For navigation equipment, if the GPS positioning deviation is found, the calibration coefficient is corrected according to the deviation value and the characteristics of the equipment. Assume that the calibration coefficient correction formula is ,in is the corrected calibration coefficient, is the calibration coefficient before correction, It is a correction factor determined according to the type of navigation equipment and accuracy requirements. The GPS positioning deviation value is used to improve the positioning accuracy of navigation equipment. Regarding power equipment, when an imbalance in power load is detected, a load balancing algorithm is used to redistribute the load. For example, a portion of the power from a high-load device can be allocated to a low-load device according to a certain ratio. This ensures the overall stable operation of the power equipment and avoids equipment failures caused by local overloads. By generating and executing these instructions for abnormal events of different priorities, various abnormal situations during unmanned tower operations can be effectively addressed, ensuring stable and reliable operation.

[0091] 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," "includes," 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.

[0092] 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 unmanned tower operation status monitoring system based on big data, characterized in that: include: A data acquisition unit, which is used to collect communication link status parameters of the unmanned tower, meteorological sensor monitoring parameters, navigation equipment operating parameters, power equipment load parameters, and aircraft dynamic trajectory parameters in real time through a stream data processing framework; a multi-source data fusion unit, configured to perform cross-domain feature extraction and heterogeneous data fusion on the communication link status parameters, meteorological sensor monitoring parameters, navigation equipment operating parameters, power equipment load parameters, and aircraft dynamic trajectory parameters through a federated learning algorithm to generate a multi-modal joint feature matrix; A dynamic modeling unit, configured to construct an unmanned tower operation state transition probability graph based on a dynamic Bayesian network and taking the multimodal joint feature matrix as input, and output a real-time state evolution sequence; A real-time anomaly detection unit, configured to calculate a time series anomaly score for the real-time state evolution sequence using an isolation forest algorithm and an LSTM autoencoder to generate an anomaly confidence level; An active control instruction generation unit, the active control instruction generation unit being configured to generate a device control instruction set based on the abnormality confidence level and in combination with a reinforcement learning strategy network; The execution steps of the real-time anomaly detection unit include: Dividing the real-time state evolution sequence into time window segments, and extracting statistical feature vectors within each time window; Performing reconstruction error calculation on the statistical feature vector using the LSTM autoencoder to obtain a first anomaly score; Performing path length deviation calculation on the statistical feature vector using the isolation forest algorithm to obtain a second anomaly score; Performing weighted fusion on the first anomaly score and the second anomaly score to generate the anomaly confidence score; The execution steps of the data acquisition unit include: Receive signal strength, bit error rate, and link delay data from the communication link status parameters in real time via the Apache Kafka stream processing platform and the communication protocol interface of the unmanned tower; Collecting wind speed, visibility and pressure gradient data among the meteorological sensor monitoring parameters through distributed sensor nodes; Extracting gyroscope drift value, GPS positioning deviation and radar scanning frequency data from the navigation device operating parameters through the device log parsing engine; The execution steps of the multi-source data fusion unit also include: Build a global federated learning model, where each tower device acts as a local client and uploads encrypted local feature gradients; Performing secure weighted averaging on the local feature gradients through an aggregation server to update the global model parameters; The updated global model parameters are sent to each client, and the multimodal joint feature matrix is ​​iteratively generated.

2. The system according to claim 1, wherein The execution steps of the dynamic modeling unit include: Defining node variables of the dynamic Bayesian network based on historical operating data, including communication link stability, meteorological interference level, navigation precision reduction factor, power load fluctuation index, and aircraft trajectory offset; Calculating the conditional probability distribution of the node variables by using a Markov chain Monte Carlo method; The edge weights of the state transition probability graph are updated based on the real-time data stream to generate the real-time state evolution sequence.

3. The system according to claim 2, wherein: The execution steps of the real-time anomaly detection unit further include: The hidden layer dimension of the LSTM autoencoder is set to 30% of the input feature dimension, and a Dropout layer is used to prevent overfitting; An adaptive subtree splitting strategy is introduced into the isolation forest algorithm to dynamically adjust the abnormal segmentation threshold according to data distribution.

4. The system according to claim 1, wherein: The execution steps of the active control instruction generating unit include: Constructing a state space of a reinforcement learning policy network, including the anomaly confidence level, the equipment health index, and the environmental risk level; The reward function is defined as the entropy reduction of the state transfer matrix after device regulation. The policy network parameters are iteratively updated through a proximal policy optimization algorithm, and the device control instruction set is output.

5. The system according to claim 4, wherein: The step of generating the device control instruction set further includes: Generate emergency shutdown instructions for high-priority abnormal events and trigger redundant equipment switching protocols; Generate dynamic parameter compensation instructions for medium and low priority abnormal events, including communication link power adjustment, navigation equipment calibration coefficient correction and power load balancing distribution.

6. The system according to claim 1, wherein: Also includes: The knowledge graph reasoning unit is used to build an unmanned tower equipment association graph based on a graph neural network and extract key node influencing factors; The execution steps of the knowledge graph reasoning unit include: Mapping the communication link status parameters, navigation equipment operating parameters and power equipment load parameters into graph nodes; Calculate the dependency strength between nodes through the graph attention mechanism to generate the key node influence factor; The key node influencing factors are input into the dynamic modeling unit to optimize the state transition probability graph.

7. An electronic device, characterized in that: include: A memory for storing a computer program for the unmanned tower operation status monitoring system based on big data according to any one of claims 1 to 6; The processor is used to execute the computer program to realize real-time monitoring and control of the operating status of the unmanned tower.

Images