Surveying and mapping data cloud storage and intelligent operation and maintenance management system and method

Abstract

The invention discloses a surveying and mapping data cloud storage and intelligent operation and maintenance management system and method in the technical field of surveying and mapping data storage, and the method comprises the steps: collecting on-site multi-dimensional time sequence data including an equipment state, an environment parameter and a task progress in real time, and calculating a comprehensive abnormality triggering index; when it is monitored that the comprehensive exception triggering index exceeds the early warning threshold value of the initial fault prediction model, it is judged that a potential unknown exception event exists, high-frequency special data collection is automatically triggered, and a high-time-resolution data fragment is generated; and marking the high-time-resolution data fragment as a to-be-verified abnormal sample, and uploading the to-be-verified abnormal sample to a cloud server cluster. Through core links such as edge real-time acquisition, self-adaptive threshold triggering, online incremental learning and closed-loop experience feedback, the operation and maintenance system has the capabilities of actively identifying'undefined abnormal working conditions', dynamically evolving a prediction model and driving a field control strategy.

Description

Technical Field

[0001] This invention relates to a cloud storage and intelligent operation and maintenance management system and method for surveying and mapping data, belonging to the field of surveying and mapping data storage technology. Background Technology

[0002] With the popularization of distributed surveying terminals, unmanned surveying platforms and high-precision positioning sensor networks, data streams are showing a trend of high-frequency sampling, high-dimensional structure and coupling with sudden anomalies. Traditional static monitoring systems and fixed threshold alarm mechanisms are no longer able to support real-time stable operation across regions, nodes and task loads. The cloud and edge collaborative full life cycle intelligent operation and maintenance system is gradually becoming an important direction for industry infrastructure construction.

[0003] Existing cloud-based surveying and mapping data storage and maintenance management systems lack proactive identification capabilities, dynamic modeling capabilities, and continuous evolution mechanisms when facing unknown or unpredictable abnormal operating conditions. This leads to problems such as delayed system response to sudden faults, insufficient accuracy in fault tracing, and models gradually becoming ineffective as the environment changes. Specifically, existing technologies mainly rely on fixed thresholds and known fault modes for alarms and scheduling, which cannot effectively handle unpredictable abnormal characteristics in equipment or data streams. They also struggle to achieve a closed-loop evolution from anomaly detection and model learning to intelligent decision-making, thus limiting the system's real-time stability and long-term reliability in complex field environments. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a cloud storage and intelligent operation and maintenance management system and method for surveying and mapping data. Through core links such as real-time edge acquisition, adaptive threshold triggering, online incremental learning and closed-loop experience feedback, the operation and maintenance system has the ability to actively identify "undefined abnormal working conditions", dynamically evolve prediction models and drive on-site control strategies. It fundamentally breaks through the passive response logic of traditional host-based operation and maintenance, and achieves a comprehensive improvement in response speed, fault tracing accuracy and continuous model effectiveness.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0006] In a first aspect, the present invention provides a method for cloud storage and intelligent operation and maintenance management of surveying and mapping data, comprising:

[0007] Real-time acquisition of multi-dimensional time-series data from the field, including equipment status, environmental parameters, and task progress, and calculation of comprehensive anomaly triggering indicators;

[0008] When the comprehensive anomaly triggering index is detected to exceed the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event and automatically triggers high-frequency special data collection to generate high-temporal-resolution data fragments.

[0009] The high temporal resolution data fragments are marked as abnormal samples to be verified and uploaded to the cloud server cluster;

[0010] The system receives the abnormal sample to be verified and compares it with historical normal operating data through an online learning engine. Only when a new abnormal pattern is confirmed, the abnormal sample to be verified is converted into a confirmed abnormal sample and a temporary incremental learning task is generated.

[0011] The online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal samples, generating an enhanced fault prediction model.

[0012] The enhanced fault prediction model is applied to the current mapping task flow that generates abnormal events to perform real-time monitoring and prediction, and output operation and maintenance decision suggestions.

[0013] Based on the aforementioned operation and maintenance decision recommendations, on-site operation and maintenance operations are performed, and the initial fault prediction model is continuously fine-tuned and its structure optimized using closed-loop data from the entire process of anomaly discovery to decision execution.

[0014] Furthermore, the calculation formula for the comprehensive anomaly triggering index is as follows:

[0015]

[0016] in, This is a comprehensive anomaly triggering index, where N is the total number of feature value categories. For a moment Next Class feature values, This represents the steady-state average value of historical normal operating conditions. These are weighting coefficients that are dynamically adjusted based on on-site fluctuations.

[0017] Furthermore, the real-time sampling frequency of the high-frequency specialized data acquisition is:

[0018]

[0019] in, The sampling frequency is the real-time sampling frequency, and t is the time interval. This is the default sampling frequency. This is the sensitivity amplification factor. To comprehensively analyze abnormal triggering indicators, This is the warning threshold.

[0020] Furthermore, the high temporal resolution data fragments are marked as anomalous samples to be verified and uploaded to a cloud server cluster. This includes: performing multidimensional compression and keyframe preservation strategies on the high temporal resolution data fragments; and introducing an anomaly significance scoring function based on temporal change rate and feature gradient during the processing, the expression of which is:

[0021]

[0022] in, This represents the anomaly significance scoring function value. This is a comprehensive abnormal triggering index, where t is time. Used to characterize abnormal rates of change The feature vector within the current time slice. Used to characterize the fluctuation amplitude of multidimensional features and A coefficient used to coordinate the influence of the rate of change and the magnitude of change on the weighting.

[0023] Furthermore, the process involves receiving the abnormal sample to be verified and comparing it with historical normal operating condition data using an online learning engine. This includes: the online learning engine employing a pattern confirmation algorithm based on feature distribution offset to calculate the distribution offset index between the abnormal sample and the normal sample. When the distribution offset index exceeds the system's adaptive confirmation threshold, the sample is formally converted into a confirmed abnormal sample, and the corresponding time-series feature contour, threshold breach path, and abnormal evolution trend curve are recorded to form a reproducible behavioral label. The formula for calculating the distribution offset index is:

[0024]

[0025] Where D is the distribution shift index, and n is the total dimension of the eigenvalues. The first abnormal sample to be verified Dimensional feature representation value, This represents the steady-state average value of historical normal operating conditions. The characteristic standard deviation under historical normal operating conditions. This is an abnormal sensitivity index. A smoothing factor to prevent the denominator from approaching zero.

[0026] Furthermore, the online learning engine automatically generates temporary incremental learning tasks based on the behavioral labels and evaluates the training priorities using the following formula:

[0027]

[0028] Where P is the training priority score. Here, D is the task scheduling coefficient, and D is the distribution offset exponent. The mean of the significance of anomalies.

[0029] Furthermore, the online learning engine, based on the confirmed abnormal samples, performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model, including: updating the parameters of the initial fault prediction model based on a feature contribution and anomaly impact weight allocation strategy, expressed as:

[0030]

[0031] in, This indicates the set of parameters involved in this round of updates. To enhance the model's prediction loss for confirmed outliers, The model parameters to be optimized are... The key weight parameters of the initial model before training are KL(), which represents the KL divergence operator. This represents the output distribution of the initial fault prediction model before the update. This represents the output distribution of the updated enhanced fault prediction model. These are the regularization weight coefficients. These are the distribution constraint weight coefficients.

[0032] Secondly, the present invention provides a cloud storage and intelligent operation and maintenance management system for surveying and mapping data, comprising:

[0033] Edge data proxy module: used to collect multi-dimensional time-series data of the field in real time, including equipment status, environmental parameters and task progress, and calculate comprehensive anomaly trigger index; when the comprehensive anomaly trigger index exceeds the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event, and high-frequency special data collection is automatically triggered to generate high-temporal resolution data fragments; the high-temporal resolution data fragments are marked as anomaly samples to be verified and uploaded to the cloud server cluster;

[0034] The core module of intelligent operation and maintenance (O&M) is used to receive the anomaly samples to be verified, compare and analyze them with historical normal operating condition data through an online learning engine, and only when a new anomaly pattern is confirmed, convert the anomaly samples to be verified into confirmed anomaly samples and generate temporary incremental learning tasks. Based on the confirmed anomaly samples, the online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model to generate an enhanced fault prediction model. The enhanced fault prediction model is applied to the mapping task flow that currently generates the anomaly event for real-time monitoring and prediction, and outputs O&M decision suggestions. Based on the O&M decision suggestions, on-site O&M operations are executed, and the initial fault prediction model is continuously subjected to incremental parameter fine-tuning and structural optimization using closed-loop data from the entire process from anomaly discovery to decision execution.

[0035] Thirdly, the present invention provides a cloud storage and intelligent operation and maintenance management device for surveying and mapping data, including a processor and a storage medium;

[0036] The storage medium is used to store instructions;

[0037] The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0038] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0039] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0040] I. This solution utilizes the real-time high-frequency data acquisition and dynamic threshold triggering mechanism of the edge data proxy module to achieve immediate identification and recording of unknown abnormal operating conditions. This frees the operation and maintenance system from reliance on fixed alarm boundaries, enabling it to autonomously discover new abnormal patterns and improving the response speed to sudden failures. Furthermore, the online learning engine performs distribution offset confirmation, saliency quantification, and incremental learning task construction on the abnormal samples to be validated. This allows the system to self-update its model under constantly emerging new abnormal conditions, avoiding the problem of static prediction models gradually failing with environmental changes, thus maintaining long-term prediction accuracy and stability.

[0041] Second, the enhanced fault prediction model in this solution completes local parameter adjustments and structural optimizations without stopping global operation and maintenance services, enabling online model evolution. This avoids the computational cost overload caused by full retraining while ensuring that historical normal operation data is not forgotten, maintaining the continuous reliability of intelligent operation and maintenance decisions. Simultaneously, through a fusion inference mechanism of prediction probability, anomaly intensity, and critical failure time, the model's inference results are transformed into directly executable operation and maintenance adjustment commands and tiered early warning strategies. This ensures that prediction goes beyond alarm output and effectively drives on-site control and management, forming an intelligent decision-making chain with real-time guidance. Furthermore, the closed-loop feedback mechanism forms a structured, accumulated knowledge base from anomaly triggering, model inference, decision execution, and effect writing. Combined with long-term retention of weight calculations and dynamic experience filtering strategies, this allows the system to generate orderly, reusable, and verifiable experience assets during continuous operation, achieving self-growth and sustainable evolution of intelligent operation and maintenance. Attached Figure Description

[0042] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0043] Figure 1This is a flowchart illustrating a cloud-based method for surveying and mapping data storage and intelligent operation and maintenance management, as provided in Embodiment 1 of the present invention. Detailed Implementation

[0044] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0045] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0046] Example 1:

[0047] Please see Figure 1 This embodiment proposes a method for cloud storage and intelligent operation and maintenance management of surveying and mapping data, including the following steps:

[0048] Step one involves real-time collection of multi-dimensional time-series data from the surveying terminal, including equipment status, environmental parameters, and task progress, via an edge data proxy module deployed on the surveying terminal. When the edge data proxy module detects an operational deviation in the surveying equipment or data stream that exceeds the warning threshold of the initial fault prediction model, it immediately identifies it as a potential unknown anomaly and automatically triggers high-frequency specialized data collection. It should be noted that after continuously collecting multi-dimensional time-series data from the field in Step one, the edge data proxy module inputs this data into the initial fault prediction model to generate a real-time health score and determines the deviation based on the system's predefined dynamic warning threshold. To improve the sensitivity and robustness of the determination, the module constructs a multi-dimensional deviation measurement function for equipment status variables, environmental variables, and task execution variables, and combines this with a real-time noise adaptive weighting mechanism to form a comprehensive anomaly triggering index. This index is calculated using the following formula:

[0049]

[0050] in, This is a comprehensive anomaly triggering index, where N is the total number of feature value categories. For a moment Next Class feature values, This represents the steady-state average value of historical normal operating conditions. These are weighting coefficients that are dynamically adjusted based on on-site fluctuations. The purpose of this formula is to highlight key feature shifts and weaken background noise, so that the judgment results are both real-time and statistically significant.

[0051] when Exceeding the warning threshold output by the initial fault prediction model At this time, the system generates a potential unknown abnormal event label and triggers a high-frequency sampling mode for the abnormality. This sampling frequency has real-time self-adjustment capability and is updated according to the following formula:

[0052]

[0053] in, The sampling frequency is the real-time sampling frequency, and t is the time interval. This is the default sampling frequency. This is the sensitivity amplification factor. To comprehensively analyze abnormal triggering indicators, This serves as the warning threshold. This formula ensures that the data collection frequency is higher the deviation from normal operating conditions, thereby capturing the characteristics of the entire abnormal evolution process. Finally, the edge data proxy module records the entire process described above and, after generating high-temporal-resolution data segments, prepares to proceed to step two, the data labeling and uploading process.

[0054] Step two: The edge data proxy module marks the high-temporal-resolution, multi-dimensional time-series data segments collected around the potential unknown anomaly as anomaly samples to be verified, and simultaneously uploads them to the cloud server cluster. It should be noted that in step two, the edge data proxy module, based on the potential unknown anomaly event tags generated in step one, segments and stores the high-frequency sampling data after triggering the event according to time windows, and constructs a multi-dimensional data structure including device state sequences, environmental disturbance sequences, and model bias sequences to ensure the continuity of anomaly features in the time dimension. To avoid redundant data causing transmission burden, the module performs multi-dimensional compression and keyframe preservation strategies on this data segment. During processing, an anomaly significance scoring function based on the temporal change rate and feature gradient is introduced, the expression of which is:

[0055]

[0056] in, This represents the anomaly significance scoring function value. This is a comprehensive abnormal triggering index, where t is time. Used to characterize abnormal rates of change The feature vector within the current time slice. Used to characterize the fluctuation amplitude of multidimensional features and This is a coefficient used to coordinate the influence of the rate and magnitude of change. The purpose of this formula is to identify the most representative outlier data segments from continuous time-series data, ensuring that the data labeled as outliers to be verified has high information density and interpretability. Subsequently, the module... The distributed data segment priority sorting is completed, high-priority segments are assigned explicit anomaly markers, and upload parameters are encapsulated in a metadata structure, including sampling frequency, trigger threshold out-of-bounds range, on-site task number, and collection timestamp. To further ensure sample integrity, the edge data proxy module adopts a differentiated upload strategy based on the formula... Automatically adjust the upload channel bandwidth usage ratio, among which Indicates upload priority weight. This is the communication adjustment coefficient. This represents the anomaly saliency scoring function value. The formula aims to ensure that segments with high anomaly saliency are uploaded first, while regular background data does not affect the cloud processing rhythm. Finally, after the data segments have been marked and transmitted in layers, they enter the verification stage of the intelligent operation and maintenance core module in step three after being uploaded to the cloud server cluster.

[0057] Step three: After receiving the abnormal sample to be verified, the intelligent operation and maintenance core module calls its built-in online learning engine to compare and analyze the sample with real-time historical normal operating condition data. If it is confirmed that the sample represents a new abnormal pattern, the sample is transformed into a confirmed abnormal sample, and a corresponding temporary incremental learning task is immediately generated. It should be noted that in step three, after receiving data marked as an abnormal sample to be verified, the intelligent operation and maintenance core module uses the online learning engine to retrieve real-time historical normal operating condition data under the corresponding task scenario to construct a comparable data alignment space, and generates a comparable feature vector sequence through a unified scale mapping and feature reconstruction method. To ensure the discriminative and adaptive nature of the identification process, the engine enables a pattern confirmation algorithm based on feature distribution offset and calculates the distribution offset index between the abnormal sample and the normal sample, the expression of which is as follows:

[0058]

[0059] Where D is the distribution shift index, and n is the total dimension of the eigenvalues. The first abnormal sample to be verified Dimensional feature representation value, This represents the steady-state average value of historical normal operating conditions. The characteristic standard deviation under historical normal operating conditions. This is an abnormal sensitivity index. A smoothing factor is used to prevent the denominator from approaching zero, and to avoid ratio overflow when the eigenvalues ​​approach a stable state. The formula aims to quantify whether outlier samples belong to independent patterns outside the known probability distribution. When When the system's adaptive confirmation threshold is exceeded, the module formally transforms the sample into a confirmed anomalous sample and records the corresponding temporal feature contour, threshold breach path, and anomalous evolution trend curve to form a reproducible behavioral label. Subsequently, the online learning engine automatically generates a temporary incremental learning task based on this label and evaluates the training priority using the following formula:

[0060]

[0061] Where P is the training priority score. Here, D is the task scheduling coefficient, and D is the distribution offset exponent. This represents the mean significance of anomalies. The purpose of this expression is to determine the urgency of the learning task in the model evolution queue, allowing key anomaly information to quickly enter the model correction channel. After completing the above process, the system enters the incremental model adjustment phase in step four.

[0062] Step four: Based on the confirmed anomaly samples, the online learning engine performs local, incremental parameter fine-tuning and structural optimization on the initial fault prediction model built into the intelligent operation and maintenance core module, generating an enhanced fault prediction model with specific scenario adaptability. This process does not interrupt the system's global prediction service. It should be noted that in step four, after receiving the anomaly labels confirmed in step three, the online learning engine does not completely retrain the global model structure. Instead, it performs a local incremental optimization process based on the gradient-sensitive regions of the model to ensure that the model maintains its ability to distinguish between historical normal operating conditions and existing anomaly categories when learning new anomaly patterns, and does not produce a forgetting effect. In this stage, model parameter updates follow a feature contribution and anomaly impact weight allocation strategy. In-situ updates are performed by dynamically selecting a subset of parameters to be adjusted. The online training strategy follows the following objective function:

[0063]

[0064] in, This indicates the set of parameters involved in this round of updates. To enhance the model's prediction loss for confirmed outliers, The model parameters to be optimized are... The key weight parameters of the initial model before training are KL(), which represents the KL divergence operator. This represents the output distribution of the initial fault prediction model before the update. This represents the output distribution of the updated enhanced fault prediction model. These are the regularization weight coefficients. The distribution constraint weights are responsible for balancing the model's stability and adaptability. The purpose of this formula is to ensure that the enhanced model maintains continuity with known patterns when learning new anomalous behaviors, thus avoiding a decline in overall prediction performance due to local corrections. After completing this update process, the online learning engine writes the key discrimination nodes and anomalous feature embedding vectors of the enhanced fault prediction model into the continuous evolution structure and deploys them synchronously in the current operating environment. This allows the model to take over prediction tasks related to the anomalous scenario in real time without pausing the global prediction service. As the enhanced model enters the running state, the system logic naturally progresses to step five: real-time monitoring and intelligent decision-making.

[0065] Step five: The enhanced fault prediction model is immediately applied to the current mapping task flow that generated the anomaly, performing real-time monitoring and prediction of subsequent data, and outputting operational decision recommendations for this new anomaly, including equipment adjustment instructions or risk warning levels. It should be noted that in step five, the enhanced fault prediction model takes over the online inference of the current mapping task flow, and the model outputs the probability of a critical fault occurring in the near future for each time window. Compared with the predicted critical failure time Meanwhile, the system continues to utilize the mean of the anomalies obtained in steps two and three. With distribution offset index As an exogenous risk indicator, it enables the generation of operation and maintenance decisions after the fusion of multi-source information. To quantify probabilistic predictions and anomaly intensity into actionable operation and maintenance recommendations, a decision confidence index is defined. The formula used to characterize the urgency of requiring an immediate response to this abnormal event is as follows:

[0066]

[0067] in, For decision confidence, This represents the probability of a critical failure occurring in the near future. The mean of the anomaly significance is given, and D is the distribution shift index. This is a smoothing scaling factor used to suppress excessive amplification at extreme anomalies. The formula aims to combine the model's probability output and the intensity of anomalies detected at the edges into a single comparable risk measure; a larger value indicates a greater need for immediate action. This is based on decision confidence. Compared with the critical failure time predicted by the enhanced model Define the intensity of operation and maintenance actions As a direct quantification of the adjustment range and intervention priority of on-site equipment, the calculation formula is:

[0068]

[0069] in, To increase the intensity of maintenance operations, For decision confidence, The critical failure time predicted by the enhanced model. This is a time smoothing term used to avoid abnormal spikes in motion intensity caused by short-term noise. The formula's effect is to output greater motion intensity when the predicted probability is high, the anomaly is significant, and the expected failure time is short, thereby driving more intense control commands and higher-level risk warnings. The system will... The risk level is mapped to a discrete risk level. High risk corresponds to proactively issuing adjustment instructions and increasing the sampling frequency; medium risk corresponds to suggesting manual intervention and enabling enhanced monitoring; and low risk corresponds to maintaining regular monitoring and recording samples for future review. The scaling factor is used to calculate the control quantity of specific equipment. The control quantity increases smoothly in increments according to the feedback loop until the risk is reduced to an acceptable range. After the decision is generated, the operation and maintenance instructions are issued to the field execution unit through automatic or auxiliary processes, and the execution results are recorded synchronously with the newly generated high-resolution data to enter the closed-loop feedback and storage process in step six.

[0070] Step Six: Based on the operational decision recommendations, the system automatically or assisted management personnel in performing on-site operational operations. The closed-loop data from the entire process of anomaly detection to decision execution is then fed back to the distributed object storage module as a new round of system experience for continuous model evolution. It should be noted that in Step Six, after completing on-site execution, the system uniformly collects the action process, response latency, equipment status change trends, and final execution results, establishing a one-to-one data link with the operational decision results generated in Step Five. This creates a closed-loop structure for the anomaly event, from detection, prediction, intervention, to backtracking. To ensure the long-term value of the feedback data entering the distributed object storage module, the system generates feedback authenticity metrics based on the execution results. Its definition is as follows:

[0071]

[0072] in, To provide feedback on the metrics for authenticity, For decision confidence, To increase the intensity of maintenance operations, This refers to the actual action response range after the on-site execution unit completes the control. A smoothing factor is used to prevent the denominator from approaching zero. The purpose of this formula is to quantify whether the execution result matches the expected behavior of the decision output. A value closer to 1 indicates a high degree of consistency between the execution behavior and the model's inference, reflecting the model's correctness in recognizing the abnormal pattern. Based on this, the system further calculates the long-term retention weights of the samples. The expression used to determine the storage level of this event and the priority of future training calls is as follows:

[0073]

[0074] in, To retain weights for samples over the long term, To provide feedback on the metrics for authenticity, The mean of the significance of the anomalies. For time smoothing term, This represents the critical failure time predicted by the enhanced model. The purpose of this expression is to prioritize and reproduce significant and predictive anomalous events, while gradually reducing the weight of noisy or sporadic events during long-term model evolution. Ultimately, this closed-loop data and weight labels, along with execution metadata, are written to a distributed object storage module, forming learning samples for the next round of model training, parameter calibration, and dynamic baseline updates. The system's operational logic then returns to the continuous monitoring phase, thus constituting a complete, accumulative, and continuously evolving intelligent operation and maintenance lifecycle.

[0075] In summary, this solution provides a novel cloud-based cloud storage and intelligent operation and maintenance management solution for surveying and mapping data that can proactively identify, dynamically learn, and continuously evolve. It breaks through the reliance of traditional operation and maintenance models on preset faults and fixed thresholds. Through real-time edge monitoring and adaptive triggering mechanisms, it achieves immediate detection of unknown abnormal operating conditions. Furthermore, by leveraging a cloud-based online incremental learning engine, the system can autonomously model and optimize new abnormal patterns, ultimately forming a complete self-evolving capability from anomaly detection, intelligent prediction, decision execution to closed-loop feedback. This comprehensively improves the system's real-time response speed, fault warning accuracy, long-term operational stability, and intelligence level in complex field environments.

[0076] Example 2:

[0077] A cloud-based surveying and mapping data storage and intelligent operation and maintenance management system, which can implement the cloud-based surveying and mapping data storage and intelligent operation and maintenance management method described in Embodiment 1, includes:

[0078] Edge data proxy module: used to collect multi-dimensional time-series data of the field in real time, including equipment status, environmental parameters and task progress, and calculate comprehensive anomaly trigger index; when the comprehensive anomaly trigger index exceeds the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event, and high-frequency special data collection is automatically triggered to generate high-temporal resolution data fragments; the high-temporal resolution data fragments are marked as anomaly samples to be verified and uploaded to the cloud server cluster;

[0079] The core module of intelligent operation and maintenance (O&M) is used to receive the anomaly samples to be verified, compare and analyze them with historical normal operating condition data through an online learning engine, and only when a new anomaly pattern is confirmed, convert the anomaly samples to be verified into confirmed anomaly samples and generate temporary incremental learning tasks. Based on the confirmed anomaly samples, the online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model to generate an enhanced fault prediction model. The enhanced fault prediction model is applied to the mapping task flow that currently generates the anomaly event for real-time monitoring and prediction, and outputs O&M decision suggestions. Based on the O&M decision suggestions, on-site O&M operations are executed, and the initial fault prediction model is continuously subjected to incremental parameter fine-tuning and structural optimization using closed-loop data from the entire process from anomaly discovery to decision execution.

[0080] Example 3:

[0081] This invention also provides a cloud storage and intelligent operation and maintenance management device for surveying and mapping data, which can realize the cloud storage and intelligent operation and maintenance management method for surveying and mapping data described in Embodiment 1, including a processor and a storage medium;

[0082] The storage medium is used to store instructions;

[0083] The processor is configured to operate according to the instructions to perform the steps of the following method:

[0084] Real-time acquisition of multi-dimensional time-series data from the field, including equipment status, environmental parameters, and task progress, and calculation of comprehensive anomaly triggering indicators;

[0085] When the comprehensive anomaly triggering index is detected to exceed the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event and automatically triggers high-frequency special data collection to generate high-temporal-resolution data fragments.

[0086] The high temporal resolution data fragments are marked as abnormal samples to be verified and uploaded to the cloud server cluster;

[0087] The system receives the abnormal sample to be verified and compares it with historical normal operating data through an online learning engine. Only when a new abnormal pattern is confirmed, the abnormal sample to be verified is converted into a confirmed abnormal sample and a temporary incremental learning task is generated.

[0088] The online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal samples, generating an enhanced fault prediction model.

[0089] The enhanced fault prediction model is applied to the current mapping task flow that generates abnormal events to perform real-time monitoring and prediction, and output operation and maintenance decision suggestions.

[0090] Based on the aforementioned operation and maintenance decision recommendations, on-site operation and maintenance operations are performed, and the initial fault prediction model is continuously fine-tuned and its structure optimized using closed-loop data from the entire process of anomaly discovery to decision execution.

[0091] Example 4:

[0092] This invention also provides a computer-readable storage medium that can implement the cloud storage and intelligent operation and maintenance management method for surveying and mapping data described in Embodiment 1. The medium stores a computer program that, when executed by a processor, performs the steps of the following method:

[0093] Real-time acquisition of multi-dimensional time-series data from the field, including equipment status, environmental parameters, and task progress, and calculation of comprehensive anomaly triggering indicators;

[0094] When the comprehensive anomaly triggering index is detected to exceed the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event and automatically triggers high-frequency special data collection to generate high-temporal-resolution data fragments.

[0095] The high temporal resolution data fragments are marked as abnormal samples to be verified and uploaded to the cloud server cluster;

[0096] The system receives the abnormal sample to be verified and compares it with historical normal operating data through an online learning engine. Only when a new abnormal pattern is confirmed, the abnormal sample to be verified is converted into a confirmed abnormal sample and a temporary incremental learning task is generated.

[0097] The online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal samples, generating an enhanced fault prediction model.

[0098] The enhanced fault prediction model is applied to the current mapping task flow that generates abnormal events to perform real-time monitoring and prediction, and output operation and maintenance decision suggestions.

[0099] Based on the aforementioned operation and maintenance decision recommendations, on-site operation and maintenance operations are performed, and the initial fault prediction model is continuously fine-tuned and its structure optimized using closed-loop data from the entire process of anomaly discovery to decision execution.

[0100] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative and not exhaustive. All modifications within the scope of this invention or its equivalents are included in this invention.

[0101] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0102] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0103] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0104] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for cloud storage and intelligent operation and maintenance management of surveying and mapping data, characterized in that, include: Real-time acquisition of multi-dimensional time-series data from the field, including equipment status, environmental parameters, and task progress, and calculation of comprehensive anomaly triggering indicators; When the comprehensive anomaly triggering index is detected to exceed the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event and automatically triggers high-frequency special data collection to generate high-temporal-resolution data fragments. The high temporal resolution data fragments are marked as abnormal samples to be verified and uploaded to the cloud server cluster; The system receives the abnormal sample to be verified and compares it with historical normal operating data through an online learning engine. Only when a new abnormal pattern is confirmed, the abnormal sample to be verified is converted into a confirmed abnormal sample and a temporary incremental learning task is generated. The online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal samples, generating an enhanced fault prediction model. The enhanced fault prediction model is applied to the current mapping task flow that generates abnormal events to perform real-time monitoring and prediction, and output operation and maintenance decision suggestions. Based on the aforementioned operation and maintenance decision recommendations, on-site operation and maintenance operations are performed, and the initial fault prediction model is continuously fine-tuned and its structure optimized using closed-loop data from the entire process of anomaly discovery to decision execution.

2. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 1, characterized in that, The formula for calculating the comprehensive anomaly triggering index is as follows: ； in, This is a comprehensive anomaly triggering index, where N is the total number of feature value categories. For a moment Next Class feature values, This represents the steady-state average value of historical normal operating conditions. These are weighting coefficients that are dynamically adjusted based on on-site fluctuations.

3. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 1, characterized in that, The real-time sampling frequency of the high-frequency special data acquisition is: ； in, The sampling frequency is the real-time sampling frequency, and t is the time interval. This is the default sampling frequency. This is the sensitivity amplification factor. To comprehensively analyze abnormal triggering indicators, This is the warning threshold.

4. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 1, characterized in that, The high temporal resolution data segments are marked as anomalous samples to be verified and uploaded to a cloud server cluster. This includes: performing multidimensional compression and keyframe preservation strategies on the high temporal resolution data segments; and introducing an anomaly significance scoring function based on temporal change rate and feature gradient during the processing, the expression of which is: ； in, This represents the anomaly significance scoring function value. This is a comprehensive abnormal triggering index, where t is time. Used to characterize abnormal rates of change The feature vector within the current time slice. Used to characterize the fluctuation amplitude of multidimensional features and A coefficient used to coordinate the influence of the rate of change and the magnitude of change on the weighting.

5. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 1, characterized in that, The process involves receiving the abnormal sample to be verified and comparing it with historical normal operating condition data using an online learning engine. This includes: the online learning engine employing a pattern confirmation algorithm based on feature distribution offset to calculate the distribution offset index between the abnormal sample and normal samples. When the distribution offset index exceeds the system's adaptive confirmation threshold, the sample is formally converted into a confirmed abnormal sample, and the corresponding time-series feature contour, threshold breach path, and abnormal evolution trend curve are recorded to form a reproducible behavioral label. The formula for calculating the distribution offset index is: ； Where D is the distribution shift index, and n is the total dimension of the eigenvalues. The first abnormal sample to be verified Dimensional feature representation value, This represents the steady-state average value of historical normal operating conditions. The characteristic standard deviation under historical normal operating conditions. This is an abnormal sensitivity index. A smoothing factor to prevent the denominator from approaching zero.

6. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 5, characterized in that, The online learning engine automatically generates temporary incremental learning tasks based on the behavioral labels and evaluates the training priorities using the following formula: ； Where P is the training priority score. Here, D is the task scheduling coefficient, and D is the distribution offset exponent. The mean of the significance of anomalies.

7. The method for cloud storage and intelligent operation and maintenance management of surveying and mapping data according to claim 1, characterized in that, The online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal samples, including: updating the parameters of the initial fault prediction model based on a feature contribution and anomaly impact weight allocation strategy, expressed as: ； in, This indicates the set of parameters involved in this round of updates. To enhance the model's prediction loss for confirmed outliers, The model parameters to be optimized are... The key weight parameters of the initial model before training are KL(), which represents the KL divergence operator. This represents the output distribution of the initial fault prediction model before the update. This represents the output distribution of the updated enhanced fault prediction model. These are the regularization weight coefficients. These are the distribution constraint weight coefficients.

8. A cloud-based storage and intelligent operation and maintenance management system for surveying and mapping data, characterized in that, include: Edge data proxy module: used to collect multi-dimensional time-series data of the field in real time, including equipment status, environmental parameters and task progress, and calculate comprehensive anomaly trigger index; when the comprehensive anomaly trigger index exceeds the warning threshold of the initial fault prediction model, it is determined to be a potential unknown anomaly event, and high-frequency special data collection is automatically triggered to generate high-temporal resolution data fragments; the high-temporal resolution data fragments are marked as anomaly samples to be verified and uploaded to the cloud server cluster; The core module of intelligent operation and maintenance is used to receive the abnormal sample to be verified, compare and analyze it with historical normal operating condition data through an online learning engine, and only when a new abnormal mode is confirmed, convert the abnormal sample to be verified into a confirmed abnormal sample and generate a temporary incremental learning task; the online learning engine performs incremental parameter fine-tuning and structural optimization on the initial fault prediction model based on the confirmed abnormal sample to generate an enhanced fault prediction model. The enhanced fault prediction model is applied to the current mapping task flow that generates abnormal events to perform real-time monitoring and prediction, and output operation and maintenance decision suggestions. Based on the aforementioned operation and maintenance decision recommendations, on-site operation and maintenance operations are performed, and the initial fault prediction model is continuously fine-tuned and its structure optimized using closed-loop data from the entire process of anomaly discovery to decision execution.

9. A cloud storage and intelligent operation and maintenance management device for surveying and mapping data, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1 to 7.

Images