Cooperative operation and maintenance management system and method for heterogeneous low-altitude facilities

Abstract

The application relates to the technical field of operation and maintenance management, in particular to a cooperative operation and maintenance management system and method for heterogeneous low-altitude facilities, wherein the system comprises: a data unification module, which is used for template mapping of fields in an original operation state sequence, and is used for normalizing processing to generate a state vector with a unified structure, and is used for splicing the state vector to generate a standard state vector sequence; an abnormal device identification module, which is used for calculating the average value of a cooperative disturbance score in a sliding window, and is used for screening a candidate abnormal device list composed of devices meeting the conditions according to the average value; an abnormality judgment module, which is used for generating a predicted state of the candidate abnormal device list in a historical normal mode through a pre-trained model, and is used for constructing a high-trust abnormal device set through abnormality credibility scoring; and a task suggestion generation module, which is used for calculating a comprehensive priority score of the high-trust abnormal device set through a scoring function, and is used for outputting a task suggestion based on the comprehensive priority score and an actual operation and maintenance process.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management technology, and more particularly to a collaborative operation and maintenance management system and method for heterogeneous low-altitude facilities. Background Technology

[0002] With the rapid development of the low-altitude economy, the deployment of facilities such as drones, low-altitude communication base stations, and environmental monitoring sensors in urban operation support, traffic monitoring, emergency response, and logistics services continues to expand. Compared with traditional ground facilities, low-altitude facilities have a wider distribution range, more complex operating environments, and a high degree of diversity in equipment types and manufacturers, which makes operation and maintenance management face higher uncertainties and complexities. On the one hand, different types of low-altitude equipment differ significantly in data structure, communication protocols, and operating indicators. Existing operation and maintenance systems often rely on customized adaptations or single equipment models, making it difficult to achieve unified management within a single platform. On the other hand, low-altitude facilities operate in environments with wind disturbance, electromagnetic interference, high temperature, and high humidity for extended periods, resulting in frequent fluctuations in equipment status and anomalies that are obviously sporadic and interconnected. Relying solely on single equipment thresholds or local indicators can easily misjudge short-term disturbances as faults or overlook systemic risks with propagation characteristics. Furthermore, most existing technologies remain at the level of anomaly alarms or status displays, lacking the characterization of inter-equipment collaboration and a complete connection mechanism from anomaly identification to operation and maintenance handling. This leads to operation and maintenance responses often relying mainly on manual experience, resulting in low processing timeliness and resource utilization efficiency. Even with the introduction of predictive models or centralized analysis methods in some solutions, there are still common problems such as insufficient adaptability to heterogeneous equipment, limited ability to identify real anomalies, and difficulty in forming executable operation and maintenance decisions. Therefore, in the context of high-density deployment of low-altitude facilities and collaborative operation of multiple devices, how to accurately identify systemic anomalies and form an executable operation and maintenance closed loop while uniformly managing heterogeneous equipment has become a key technical problem that urgently needs to be solved in the current operation and maintenance field of low-altitude facilities. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a collaborative operation and maintenance management system and method for heterogeneous low-altitude facilities.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A collaborative operation and maintenance management system for heterogeneous low-altitude facilities includes: The data unification module is used to obtain the original operating state sequence of the device, perform template mapping on the fields in the original operating state sequence through a preset state field template, and normalize the mapped fields to generate a state vector with a unified structure. The state vector is then concatenated to generate a standard state vector sequence. The abnormal device identification module is used to calculate the collaborative residual between devices through the collaborative residual function, calculate the collaborative disturbance score of the devices based on the collaborative residual, smooth the collaborative disturbance score through a sliding window, calculate the average value of the collaborative disturbance score within the sliding window, and filter the devices that meet the conditions to form a candidate abnormal device list based on the average value. The anomaly detection module is used to predict and reconstruct the current state of the candidate abnormal device list using a pre-trained model, generate the predicted state of the candidate abnormal device list under historical normal mode, calculate the difference between the current state and the predicted state of the candidate abnormal device list, generate an anomaly confidence score, and construct a set of highly reliable abnormal devices based on the anomaly confidence score. The task suggestion generation module is used to calculate the comprehensive priority score of a set of highly reliable abnormal devices through a scoring function, and output task suggestions based on the comprehensive priority score and the actual operation and maintenance process.

[0005] Collaborative operation and maintenance management methods for heterogeneous low-altitude facilities include: The original operating state sequence of the device is obtained, and the fields in the original operating state sequence are mapped using a preset state field template. The mapped fields are then normalized to generate a state vector with a unified structure. The state vectors are then concatenated to generate a standard state vector sequence. The collaborative residual between devices is calculated using a collaborative residual function. Based on the collaborative residual, the collaborative disturbance score of the devices is calculated. The collaborative disturbance score is smoothed using a sliding window, and the average value of the collaborative disturbance score within the sliding window is calculated. Based on the average value, devices that meet the criteria are selected to form a candidate list of abnormal devices. The candidate abnormal device list is predicted and reconstructed at the current time using a pre-trained model to generate the predicted state of the candidate abnormal device list under the historical normal mode. The difference between the current state and the predicted state of the candidate abnormal device list is calculated to generate an anomaly confidence score. A set of highly reliable abnormal devices is constructed based on the anomaly confidence score. The comprehensive priority score of the set of highly reliable abnormal devices is calculated using a scoring function, and task suggestions are output based on the comprehensive priority score and the actual operation and maintenance process.

[0006] Preferably, the status field template includes power supply status, temperature status, and signal strength status.

[0007] Preferably, the parameters of the cooperative residual function include a standard Euclidean residual term, an absolute difference term, a weighting coefficient, and a structural regularization strength coefficient. The standard Euclidean residual term is used to represent the main differences between devices in the current state, the absolute difference term is used to amplify local high-amplitude deviations, and the structural regularization strength coefficient is used to adjust the degree of cooperative topology influence.

[0008] Preferably, the pre-trained model is an autoencoder structure, which includes an input layer, a single hidden layer, and an output layer. The training of the model is completed at the device level or at an edge node.

[0009] Preferably, the predicted state is generated by a prediction function, the parameters of which include the device's state vector at the previous time step, the weights and biases of the coding layer, the nonlinear activation operator, and the parameters of the decoding layer.

[0010] Preferably, the anomaly credibility score is generated by a credibility evaluation function, the parameters of which include the direct deviation of the device's current state from its own historical normal mode, the cooperative disturbance score, the average of the cooperative disturbance scores of all candidate devices at the current time point, and the adjustment coefficient.

[0011] Preferably, the parameters of the scoring function include anomaly confidence score, cooperative perturbation score, and adjustment factor.

[0012] Preferably, the task recommendations include remote command, on-site inspection, and continuous monitoring.

[0013] Preferably, after outputting task suggestions based on comprehensive priority scoring and actual operation and maintenance processes, the method further includes: The comprehensive priority score and task suggestions are structurally encapsulated and output to the task scheduling platform. The beneficial effects of this invention are as follows: This invention first establishes a unified model for the operational status of different types of low-altitude equipment, eliminating differences in data structure and numerical scale between devices and providing a consistent foundation for cross-device analysis. Based on this, by constructing a collaborative state residual structure among devices, it identifies key devices that significantly disrupt collaborative operation from a system-wide perspective, avoiding isolated judgments from a single device's perspective. Furthermore, this invention combines the historical operational characteristics of each device to determine the authenticity of candidate abnormal devices, thus distinguishing between actual device degradation and apparent anomalies caused by environmental disturbances or collaborative propagation. Finally, this invention integrates anomaly credibility with the degree of system collaborative impact to generate maintenance task suggestions that match the device type and control capabilities, achieving closed-loop management from anomaly identification to maintenance execution. Through the above technical solutions, this invention not only improves the accuracy and stability of anomaly identification for heterogeneous low-altitude facilities but also provides clear execution directions for maintenance decisions, effectively enhancing the response efficiency and actual availability of collaborative maintenance of low-altitude facilities. Attached Figure Description

[0014] Figure 1 This is a block diagram of a collaborative operation and maintenance management system for heterogeneous low-altitude facilities in a specific embodiment of the present invention; Figure 2 This is a flowchart of a collaborative operation and maintenance management method for heterogeneous low-altitude facilities in a specific embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please refer to Figure 1 As shown, this invention proposes a collaborative operation and maintenance management system for heterogeneous low-altitude facilities, including: The data unification module is used to obtain the original operating state sequence of the device, perform template mapping on the fields in the original operating state sequence through a preset state field template, and normalize the mapped fields to generate a state vector with a unified structure. The state vector is then concatenated to generate a standard state vector sequence. The abnormal device identification module is used to calculate the collaborative residual between devices through the collaborative residual function, calculate the collaborative disturbance score of the devices based on the collaborative residual, smooth the collaborative disturbance score through a sliding window, calculate the average value of the collaborative disturbance score within the sliding window, and filter the devices that meet the conditions to form a candidate abnormal device list based on the average value. The anomaly detection module is used to predict and reconstruct the current state of the candidate abnormal device list using a pre-trained model, generate the predicted state of the candidate abnormal device list under historical normal mode, calculate the difference between the current state and the predicted state of the candidate abnormal device list, generate an anomaly confidence score, and construct a set of highly reliable abnormal devices based on the anomaly confidence score. The task suggestion generation module is used to calculate the comprehensive priority score of a set of highly reliable abnormal devices through a scoring function, and output task suggestions based on the comprehensive priority score and the actual operation and maintenance process.

[0017] Please refer to Figure 2 As shown, in another aspect of the present invention, a collaborative operation and maintenance management method for heterogeneous low-altitude facilities is also proposed, including: Step 1: Obtain the original operating state sequence of the device. Map the fields in the original operating state sequence using a preset state field template, and normalize the mapped fields to generate a state vector with a unified structure. Concatenate the state vectors to generate a standard state vector sequence. Specifically, this includes: This step is used to construct a unified representation of the state of low-altitude equipment, resolving inconsistencies in data structure, indicator meaning, and numerical scale among heterogeneous devices. Different equipment types, such as drones, communication base stations, and environmental sensors, collect data fields that differ, making direct collaborative analysis in the same space impossible. This step uses a two-stage processing method—field template mapping and numerical normalization—to transform the raw collected data into a structurally unified and comparable time-series state vector, providing standard input for subsequent collaborative anomaly identification.

[0018] The input is the sequence of raw operating statuses collected by each device within a certain time period, denoted as: ; in, Indicates the first The original state sequence of each device is obtained from the device body by the edge device or embedded acquisition unit through a communication interface (such as RS485, serial port, Ethernet); This indicates the length of the time window covered by the sequence; Indicates the first The device in the The original state vector at each point in time has a field structure that varies depending on the type of device. For example, a drone may contain battery voltage, current, flight control temperature, and GPS signal strength, while an environmental sensor may contain air temperature, humidity, and power supply status.

[0019] First, template mapping is performed on the field structure. The system predefines a set of status field templates that are shared across devices. Each of them This represents a general device status attribute, such as power supply status, temperature status, signal strength status, etc. The device's original fields are mapped to this template according to their meaning; missing fields are filled with default values, such as... or .

[0020] To eliminate the problem of inconsistent numerical scales, each mapping field is normalized: ; in, Indicates the first The device at the time point Above, the first after field mapping One original index value; For the first The average value of each field in historical device data is calculated offline during the deployment period; The corresponding standard deviation; These are normalized state values, which are comparable.

[0021] The normalized results of each dimension are combined into a state vector with a unified structure: ; The final standard state vector sequence is formed by concatenating multiple time points: ; Each of them The normalized device state vector has a fixed dimension. This structure is shared by all devices. This state sequence will be used as input in the next step to construct the inter-device state residual matrix and anomaly filtering.

[0022] This solution achieves structural unification and numerical alignment of heterogeneous low-altitude equipment status data through field templates and standardization mechanisms, avoiding the complexity of modeling each equipment individually, and enabling collaborative status analysis between different equipment in the same space.

[0023] Step 2: Calculate the collaborative residuals between devices using the collaborative residual function. Based on the collaborative residuals, calculate the collaborative disturbance score of the devices. Smooth the collaborative disturbance score using a sliding window and calculate the average value of the collaborative disturbance score within the sliding window. Based on the average value, select devices that meet the criteria to form a candidate list of abnormal devices. Specifically, this includes: This step addresses the high-density deployment scenario of various types of low-altitude facilities by proposing a systematic anomaly identification method based on inter-device state coordination residual calculation. This method is used to screen candidate devices that currently cause the greatest interference to the overall system's coordinated state. Unlike traditional methods that rely on single-point threshold judgments or simple clustering, this step compares the state consistency among multiple devices and incorporates factors such as deployment area characteristics and device role differences. It introduces a structural residual regularization term to construct a more interpretable anomaly indicator system, thereby improving the identification accuracy and deployment adaptability in complex deployment environments.

[0024] The input is the standard state vector sequence output from the previous step. Each of them Indicates the first The device at the time point The standardized state vector on, As for the feature dimension, this vector structure was already based on a cross-device shared template in the previous step. The process is unified. All device data is synchronously cached by the edge gateway at a fixed sampling frequency (e.g., 5 seconds / time), and after time alignment, it is used as the input for this step.

[0025] Considering that low-altitude facilities are often deployed intensively in local areas, and some devices naturally have operational or physical connections (such as shared power supply and signal relay), if state residual comparison between devices is performed without differentiation, the linked anomaly risk caused by some devices in their functional topology may be underestimated. Therefore, this step designs a weighted collaborative residual modeling method combined with deployment structure, and introduces a structural regularization term based on traditional residual calculation, so as to capture potential anomaly sources more effectively.

[0026] First, define the current time point device and device the collaborative residual between them is , through the collaborative residual function , which is specifically: ; wherein, is a standard Euclidean residual term, representing the main difference between the current states of two devices; is an absolute difference term used to amplify local high-amplitude deviations, so as to prevent anomalies in important dimensions from being submerged by the averaging effect; is device and weight coefficient in the deployment topology, which is set according to the actual physical deployment or task coupling relationship (for example, sub-devices sharing the same power supply and under the same gateway can be set as , and those with no direct association are set as ); is a structural regularization intensity coefficient, which adjusts the influence degree of collaborative topology, and its value is usually between , which is set during the deployment period.

[0027] Through this residual structure design, it not only measures the general state difference between devices, but also injects structural knowledge and deployment logic into the residual term, so that anomalies can be significantly identified when some devices show slight abnormalities but their connection relationship is important. This design is particularly applicable to core devices with high collaborative dependence in low-altitude facilities, such as signal relay nodes and key power supply nodes.

[0028] Next, calculate device at time the total system disturbance degree is: ; wherein represents the set of adjacent devices that have deployment association or task collaborative relationship with device . The total disturbance value represents the overall collaborative disturbance degree caused by device in the current system structure, and a larger value indicates that it is more likely to be the inducement or key node of the current systemic anomaly.

[0029] To improve the stability of the results, the system... Smoothing is achieved using a sliding window, with the window length as the parameter. The average value over (e.g., 3 periods) is used as the final cooperative perturbation score, while a dynamic threshold is set. Select the one that satisfies The candidate set of devices: ; in express The mean within the sliding window, It can be dynamically adjusted based on the number of devices and risk strategies, such as setting it to all devices in the current area. mean plus Double the standard deviation.

[0030] The above mechanism not only considers the degree of deviation of the device state itself, but also integrates the structural position and collaborative value of the device in actual deployment. Through the combined effect of regularization terms and topology coupling weights, it improves the structural sensitivity and engineering adaptability of anomaly identification.

[0031] Step 3: Using a pre-trained model, predict and reconstruct the current state of the candidate abnormal device list to generate the predicted state of the candidate abnormal device list under historical normal mode. Calculate the difference between the current state and the predicted state of the candidate abnormal device list to generate an anomaly confidence score. Based on the anomaly confidence score, construct a set of highly reliable abnormal devices, specifically including: This step, based on collaborative residual screening, determines the "authenticity" of candidate anomalous devices. Its core objective is not to rediscover anomalies, but rather to distinguish between "genuine device anomalies within structural anomalies" and "apparent anomalies caused by collaborative propagation, environmental disturbances, or entanglement with neighboring devices." In the previous step, the system has already used a standard state vector sequence... A collaborative residual structure among devices was constructed, and a set of candidate abnormal devices was output. and the corresponding cooperative perturbation score These results indicate that the equipment is suspected of anomalies at the "group level," but this does not directly equate to the equipment itself having entered an abnormal operating state. This step utilizes the equipment's historical operating trajectory to further determine this issue.

[0032] The input consists of two parts. The first part is the standard state vector sequence output from step one. ,in Indicates device In time The first part is the standardized state vector, which has undergone cross-device field alignment and numerical normalization; the second part is the candidate device set output from step two. and the corresponding cooperative perturbation score This step is only for... The devices in the model perform the modeling and judgment, while the other devices do not participate in the calculation, so as to ensure that the method focuses on the "critical nodes that still need to be confirmed after a systemic anomaly".

[0033] For each candidate device The system does not introduce cross-device samples or use a global model; instead, it constructs an individualized normal state baseline solely using the device's own historical state sequences. Specifically, it selects a sub-sequence of states within a recent time window. A lightweight autoencoder structure is trained to characterize the state change patterns of the device under conditions of "no obvious faults". This autoencoder structure consists of an input layer, a single hidden layer, and an output layer. The hidden layer dimension is smaller than the state vector dimension, used to compress and extract stable operating features of the device. Model training is completed at the device level or edge nodes, using only the device's own historical data to avoid interference from differences in the operation of different devices on the judgment results.

[0034] After training, the model is used to predict and reconstruct the current state of the device. The expected state representation under the constraints of the historical normal mode is obtained through the prediction function, specifically: ; in, This is the standard state vector of the device at the previous moment; and These are the weights and biases of the coding layer, used to map the input state to a low-dimensional implicit representation; It is a nonlinear activation operator used to express the nonlinear change characteristics of the device state; and These are the parameters for the decoding layer, used to restore the implicit representation to the state space; This refers to the model's prediction of the current state under the assumption that the equipment itself is operating normally.

[0035] After obtaining the predicted state, the system compares it with the current actual state. With predicted state The degree of anomaly is measured by the difference in parameters. Considering that in low-altitude facility scenarios, some anomalies are often accompanied by sudden changes in key indicators rather than overall drift, this step introduces a collaborative perception regularization term in addition to the basic residual term. This allows the judgment result to simultaneously reflect "the degree of deviation of the equipment itself" and "its importance in system collaboration," thus constructing an anomaly credibility score. The specific credibility evaluation function is as follows: ; in, This indicates the degree of direct deviation of the device's current state from its historical normal mode; The output of step two is the device cooperative disturbance score; This represents the average of the cooperative perturbation scores of all candidate devices at the current time point, used to eliminate the impact of scale. The adjustment coefficient controls the weight of collaborative information in the judgment of anomaly authenticity. The implication of this design is that when a device not only deviates significantly from its own state but also acts as a high-disturbance node in the system's collaborative structure, its anomaly credibility will be significantly amplified, which is more in line with the actual operation and maintenance logic of "prioritizing the handling of critical node failures" in low-altitude facilities.

[0036] Through the above mechanism, this step avoids false alarms caused by judging solely based on model reconstruction errors, and also avoids the risk of directly equating collaborative residuals with equipment failures. Instead, it organically combines the two to form a joint discrimination method of "individual historical consistency + system collaborative impact." The output includes two results. The first is the anomaly confidence score for each candidate device at the current time. This score will serve as an important input for task prioritization in subsequent operation and maintenance decisions; the second is a set of highly reliable abnormal devices. The set consists of those that satisfy Devices exceeding a preset threshold constitute the target devices that need to enter the operation and maintenance execution phase.

[0037] Step 4: Calculate the comprehensive priority score of the set of highly reliable abnormal devices using a scoring function. Based on the comprehensive priority score and the actual operation and maintenance process, output task suggestions, specifically including: This step aims to further generate task types and output maintenance recommendations based on the anomaly assessment results generated in the previous stage, thereby achieving a closed-loop transformation from anomaly detection to collaborative maintenance response. Low-altitude infrastructure networks are typically deployed in complex areas such as cities, mountains, and transportation corridors. Different types of equipment (such as drones, communication base stations, and environmental sensors) have significantly different operating characteristics and controllability. Therefore, when generating task recommendations, it is necessary to consider not only the severity of the anomaly but also factors such as equipment type, remote controllability, and communication link quality to ensure that the task recommendations are feasible and engineering-practical. The input data in this step all come from the completed identification and evaluation modules, including the anomaly credibility score of the equipment. (Derived from the combination of modeling residuals and structural regularization), cooperative perturbation score (Generated by residual matrix modeling), both measures whether the device is "truly broken" and whether "its failure will affect others," and must be considered together. In addition, deployment-period registration data such as device type labels and control method support tags are introduced to assist in decision generation.

[0038] The system first targets the set of highly reliable anomalous devices confirmed in step three. Calculate the overall task priority score for each device. This score is used for scheduling and prioritizing multiple devices. It is an indicator that integrates anomaly confidence with cooperative disturbances. Its structural design takes into account both the abnormal behavior of the devices themselves and their structural impact on the system. The specific scoring function is calculated as follows: ; in For equipment The current anomaly confidence score is derived from the modeling residual calculation results of the third stage; The cooperative disturbance score of the equipment in the system structure is derived from the residual propagation graph analysis module; As an adjustment factor, it is recommended to set it to [value missing]. arrive The amplification effect of the cooperative terms is limited. This function structure guarantees that when... Larger (the equipment itself has obvious abnormalities) and When the anomaly is also large (and still spreading), the device will receive a higher priority score. In actual deployments, such devices should be prioritized for processing tasks, such as drone backhaul nodes located on main traffic arteries and low-altitude communication relay equipment.

[0039] After obtaining the task priority score, the system enters the task suggestion generation phase. The goal of task suggestions is to output a decision on "how to handle" each device and specify the task type. Based on actual operation and maintenance processes, task types mainly include three categories: remote command execution, on-site dispatch for repair, and continued monitoring and observation. The judgment logic not only relies on the score but also incorporates the control capability flags and environmental constraint parameters configured in the device attribute table. The task suggestion generation logic is as follows: ; in Indicates device The task suggestion type at the current moment, , The trigger thresholds for "serious anomalies" and "general anomalies" are respectively set (usually the upper 1 / 4 quantile and mean of the anomaly score), and the specific values ​​can be set by the system administrator before deployment or dynamically adjusted. The "remote control support flag" used in task type determination comes from the device registration information database and is configured when the device first connects to the platform. For example, most communication base stations support remote parameter adjustment, but some environmental sensors and devices deployed in mountainous areas cannot be remotely operated. Therefore, even if their anomaly score is high, they can only be suggested to generate a "field maintenance" task. In addition, for devices whose current signal strength is below the set threshold, even if they have remote capabilities, the system can predict the risk of remote failure through the fault simulation module, thereby prioritizing the dispatch of manual tasks and avoiding the operational redundancy of "delivery failure". Here, remote failure refers to not receiving a valid acknowledgment from the device side or detecting that the communication link is unreachable during the most recent command delivery process. This status is provided by the device communication monitoring module.

[0040] Practical task suggestions include not only task type It will also associate recommended execution methods. For example, for "remote command" tasks, it will generate recommended instruction types (such as restart, parameter adjustment); for "on-site maintenance" tasks, it will generate optimal response path suggestions (such as entering the scheduling queue or creating a work order) based on the equipment deployment point coordinates and the distribution of the nearest accessible maintenance personnel. All task suggestion structures will ultimately be encapsulated in the following format: This information is then passed to the task scheduling platform or front-end display system for direct use by subsequent task management modules.

[0041] For example, a drone relay node deployed under a city bridge has an anomaly credibility score of [insert score here]. The cooperative disturbance score is The corresponding calculated priority score is approximately If the device supports remote control, the system will output a "remote command" task, suggesting that parameter adjustments and log restarts be attempted first. If the device with the same score is an environmental sensor deployed in a suburban area with poor communication quality, the system will output a "field maintenance" task and provide a scheduling route. This task suggestion mechanism, which dynamically generates suggestions based on the actual scenario, device attributes, and abnormal behavior, is one of the structural innovations of this invention in operation and maintenance decision-making.

[0042] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A collaborative operation and maintenance management system for heterogeneous low-altitude facilities, characterized in that, include: The data unification module is used to obtain the original operating state sequence of the device, perform template mapping on the fields in the original operating state sequence through a preset state field template, and normalize the mapped fields to generate a state vector with a unified structure. The state vector is then concatenated to generate a standard state vector sequence. The abnormal device identification module is used to calculate the collaborative residual between devices through the collaborative residual function, calculate the collaborative disturbance score of the devices based on the collaborative residual, smooth the collaborative disturbance score through a sliding window, calculate the average value of the collaborative disturbance score within the sliding window, and filter the devices that meet the conditions to form a candidate abnormal device list based on the average value. The collaborative residual function The expression is: ； in, This is a standard Euclidean residual term, representing the main difference between the current states of the two devices; This is the absolute difference term used to amplify local high-amplitude deviations; For equipment and Weighting coefficients in the deployment topology The structural regularity strength coefficient; Let be the state vector of the i-th device. This represents the state vector of the j-th device; The anomaly detection module is used to predict and reconstruct the current state of the candidate abnormal device list using a pre-trained model, generate the predicted state of the candidate abnormal device list under historical normal mode, calculate the difference between the current state and the predicted state of the candidate abnormal device list, generate an anomaly confidence score, and construct a set of highly reliable abnormal devices based on the anomaly confidence score. The expression for the anomaly credibility score is: ； in, Scoring the credibility of anomalies. This indicates the degree of direct deviation of the device's current state from its historical normal mode; Scoring the cooperative disturbances of the equipment; This is the average of the cooperative disturbance scores of all candidate devices at the current time point; The adjustment coefficient is used to control the weight of collaborative information in the judgment of anomaly authenticity. This represents the actual state. For predicting the state; The task suggestion generation module is used to calculate the comprehensive priority score of the set of highly reliable abnormal devices through a scoring function, and output task suggestions based on the comprehensive priority score and the actual operation and maintenance process. The expression for the scoring function is: ； in For comprehensive priority scoring, For equipment Current anomaly credibility score; The score is given for the cooperative disturbance of the equipment in the system structure. It is a regulating factor.

2. A collaborative operation and maintenance management method for heterogeneous low-altitude facilities, characterized in that, include: The original operating state sequence of the device is obtained, and the fields in the original operating state sequence are mapped using a preset state field template. The mapped fields are then normalized to generate a state vector with a unified structure. The state vectors are then concatenated to generate a standard state vector sequence. The collaborative residual between devices is calculated using a collaborative residual function. Based on the collaborative residual, the collaborative disturbance score of the devices is calculated. The collaborative disturbance score is smoothed using a sliding window, and the average value of the collaborative disturbance score within the sliding window is calculated. Based on the average value, devices that meet the criteria are selected to form a candidate list of abnormal devices. The collaborative residual function The expression is: ； in, This is a standard Euclidean residual term, representing the main difference between the current states of the two devices; This is the absolute difference term used to amplify local high-amplitude deviations; For equipment and Weighting coefficients in the deployment topology The structural regularity strength coefficient; Let be the state vector of the i-th device. This represents the state vector of the j-th device; The candidate abnormal device list is predicted and reconstructed at the current time using a pre-trained model to generate the predicted state of the candidate abnormal device list under the historical normal mode. The difference between the current state and the predicted state of the candidate abnormal device list is calculated to generate an anomaly confidence score. A set of highly reliable abnormal devices is constructed based on the anomaly confidence score. The expression for the anomaly credibility score is: ； in, Scoring the credibility of anomalies. This indicates the degree of direct deviation of the device's current state from its historical normal mode; Scoring the cooperative disturbances of the equipment; This is the average of the cooperative disturbance scores of all candidate devices at the current time point; The adjustment coefficient is used to control the weight of collaborative information in the judgment of anomaly authenticity. This represents the actual state. For predicting the state; The comprehensive priority score of the set of highly reliable abnormal devices is calculated by a scoring function, and task suggestions are output based on the comprehensive priority score and the actual operation and maintenance process. The expression for the scoring function is: ； in For comprehensive priority scoring, For equipment Current anomaly credibility score; The cooperative disturbance score of the equipment in the system structure is derived from the residual propagation graph analysis module; It is a regulating factor.

3. The collaborative operation and maintenance management method for heterogeneous low-altitude facilities according to claim 2, characterized in that, The status field template includes power supply status, temperature status, and signal strength status.

4. The collaborative operation and maintenance management method for heterogeneous low-altitude facilities according to claim 2, characterized in that, The pre-trained model is an autoencoder structure, which includes an input layer, a single hidden layer, and an output layer. The training of the model is completed at the device level or at an edge node.

5. The collaborative operation and maintenance management method for heterogeneous low-altitude facilities according to claim 2, characterized in that, The predicted state is generated by a prediction function whose parameters include the device's state vector at the previous time step, the weights and biases of the coding layer, the nonlinear activation operator, and the parameters of the decoding layer.

6. The collaborative operation and maintenance management method for heterogeneous low-altitude facilities according to claim 2, characterized in that, The task recommendations include remote command, on-site inspection, and continuous monitoring.

7. The collaborative operation and maintenance management method for heterogeneous low-altitude facilities according to claim 2, characterized in that, After outputting task suggestions based on comprehensive priority scoring and actual operation and maintenance processes, the following is also included: The comprehensive priority score and task suggestions are structurally encapsulated and output to the task scheduling platform.

Images