Plateau rainy season-oriented power transmission tower landslide risk early warning analysis method and system

Abstract

The application discloses a power transmission tower landslide risk early warning analysis method and system for a plateau rainy season, relates to the field of power system safety monitoring and risk assessment, and comprises the following steps: collecting multi-source environment data; based on the multi-source environment data, extracting core characteristic factors, mapping to a distribution characteristic space, and constructing a high-resolution multi-dimensional environment modeling parameter set; acquiring tower micro-deformation monitoring data, and performing data fusion with the multi-dimensional environment modeling parameter set to obtain a time-space coupling characteristic data set; inputting the time-space coupling characteristic data set into a gradual risk evolution model, and dynamically adjusting the weight of each characteristic factor in risk determination; based on the risk evolution result after weight optimization, determining the progressive trend of the landslide risk, and generating an early warning level and a determination result. The application can improve the landslide risk identification sensitivity and early warning accuracy of the plateau power transmission tower in a complex environment.

Description

Technical Field

[0001] This application relates to the field of power system safety monitoring and risk assessment technology, and in particular to a method and system for early warning analysis of landslide risks on transmission towers in high-altitude rainy seasons. Background Technology

[0002] With the continuous growth of electricity demand and the expansion of power transmission networks, transmission lines are increasingly widely distributed in plateau regions. Plateau regions have complex terrain, significant topographic relief, and highly seasonal climate conditions. Especially during the rainy season, the risk of geological changes due to heavy rainfall and surface runoff is significantly increased. Transmission towers, as a crucial component of transmission lines, often have their foundations situated on slopes and gullies, making them susceptible to geological disasters such as soil loosening, landslides, and mudslides. Once a tower is damaged by a landslide, it can range from affecting power supply stability to causing casualties and property damage, and ultimately impacting the safe operation of the power system.

[0003] Current methods for monitoring geological hazards on power transmission towers in plateau regions primarily rely on fixed-point inspections, ground sensor deployment, or remote sensing image analysis. While these methods can detect geological movements or foundation deformation to some extent, their monitoring coverage is limited by regional climate, terrain obstacles, and the difficulty of equipment deployment, resulting in insufficient real-time and dynamic data. Furthermore, due to drastic changes in hydrological conditions during the rainy season, the stability of foundations after water immersion and erosion is difficult to accurately assess. Existing methods have limited capabilities for rapid, multi-source information fusion and dynamic risk early warning, making it difficult to achieve effective prediction and early warning before disasters occur. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for landslide risk early warning analysis of power transmission towers in plateau rainy seasons, which can improve the sensitivity and accuracy of landslide risk identification and early warning of power transmission towers in complex environments in plateau areas, and enhance the adaptive capability and long-term operational stability of the early warning system.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for early warning and analysis of landslide risks on transmission towers during the rainy season in plateau regions, the method comprising: Collect multi-source environmental data; the multi-source environmental data includes: environmental input signals and background information.

[0006] Based on multi-source environmental data, core feature factors are extracted and mapped to the distribution feature space to construct a high-resolution multidimensional environmental modeling parameter set. The core feature factors include: Terrain Roughness Index (TRI), Soil Plasticity Index (PI), Effective Pore Water Pressure (EPWP), Root Activity Index (RAI), Freeze-Thaw Cycles (FTC), and Evapotranspiration Ratio (ETR).

[0007] Micro-deformation monitoring data of the tower is acquired and fused with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupling feature dataset; the micro-deformation monitoring data includes: foundation displacement, foundation tilt angle and dynamic stress change data.

[0008] The spatiotemporal coupled feature dataset is input into the gradual risk calculus model to dynamically adjust the weight of each feature factor in risk assessment; the gradual risk calculus model includes: a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer.

[0009] Based on the risk calculation results after weight optimization, the progressive trend of landslide risk is determined, and the warning level and judgment result are generated.

[0010] Optionally, the method further includes: The warning level and judgment result are sent to the operation and maintenance emergency response system. The system automatically matches the response strategy library according to the warning level and dynamically activates the emergency response strategy. The emergency response strategy includes: on-site inspection and dispatch, risk area isolation or temporary control instructions.

[0011] Based on the aforementioned emergency response strategy, actual risk variations and environmental factor feedback data are recorded and fed back to the multidimensional environmental modeling parameter set, and the model self-updating algorithm is executed periodically.

[0012] Optionally, the warning levels include: yellow warning, orange warning, and red warning.

[0013] When the warning level is yellow, a command to dispatch drones for patrol is generated.

[0014] When the warning level is orange, an order is generated to dispatch an on-site investigation team.

[0015] When the warning level is red, an instruction is generated to activate the power outage plan and seal off the danger zone.

[0016] Optionally, multi-source environmental data can be collected, specifically including: Acquire multidimensional environmental input signals from several sensing systems; the multidimensional environmental input signals include: micro-topographic information, land physical indicators, surface hydrological changes, regional meteorological evolution, and biological indicator characteristics; the biological indicator characteristics include: changes in plant root density, evolution of surface moss coverage, and frequency of animal activity traces.

[0017] Obtain macro-environmental background information from remote big data sources; the macro-environmental background information includes: regional meteorological, geological and vegetation distribution information.

[0018] The multidimensional environmental input signal and the macroscopic environmental background information are synchronized in time and registered in space to obtain synchronized and registered data.

[0019] After the synchronized and registered data is classified, it is stored in a distributed time-series database.

[0020] Data anomaly processing is performed on the multidimensional environmental input signal and the macroscopic environmental background information to obtain anomaly-processed multi-source environmental data.

[0021] Based on the multi-source environmental data after anomaly processing, initial static weights are set.

[0022] Optionally, based on multi-source environmental data, core feature factors are extracted and mapped to a distribution feature space to construct a high-resolution multi-dimensional environmental modeling parameter set, specifically including: Standardized multi-source environmental data is obtained by performing standardized preprocessing on multi-source environmental data. Feature filtering is performed on the standardized multi-source environmental data to obtain core feature factors; Instantaneous values ​​of each characteristic factor were extracted; among them, the micro-topographic relief was calculated from laser point cloud data; the soil plasticity index was obtained by fitting the field electrical impedance measurement value to the laboratory calibration curve; the effective pore water pressure was derived from the Terzaghi effective stress principle; the biological root activity index was weighted by the root impedance change rate and the NDVI growth rate; the freeze-thaw cycle frequency was accumulated by the number of times the daily highest / lowest temperature crossed 0℃; and the surface evapotranspiration ratio was determined by the ratio of the measured value of the evapotranspiration meter to the theoretical value of potential evapotranspiration.

[0023] Based on the instantaneous values ​​of each feature factor, a multidimensional feature space is constructed to obtain the state distribution cloud map of the current monitoring area.

[0024] Based on the state distribution cloud map, a time sliding window mechanism is introduced for each feature factor to extract dynamic evolution features.

[0025] Based on the dynamic evolution characteristics, each feature factor is rasterized and assigned a spatial weight to obtain a high-resolution multidimensional environment modeling parameter set.

[0026] The multidimensional environment modeling parameter set is encapsulated into an extensible XML Schema structure.

[0027] Optionally, the multi-source environmental data can be standardized through preprocessing to obtain standardized multi-source environmental data, specifically including: The physical quantity units and data format of the multi-source environmental data are standardized to obtain unified data.

[0028] An adaptive filtering algorithm is used to remove high-frequency noise and drift interference from the unified data to obtain filtered data.

[0029] The filtered data is then normalized to obtain normalized data.

[0030] The normalized data is filtered based on a data quality scoring mechanism to obtain filtered data.

[0031] Based on the filtered data, a version management mechanism is constructed.

[0032] Based on the aforementioned management mechanism, standardized multi-source environmental data is output.

[0033] Optionally, tower micro-deformation monitoring data are acquired and fused with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupled feature dataset, specifically including: Based on the deployed multi-level micro-deformation sensing system, micro-deformation monitoring data of the tower is acquired; the multi-level micro-deformation sensing system includes: displacement sensor, MEMS inclinometer and deep inclinometer; all sensing data are aggregated through edge computing gateway, preliminarily compressed and then uploaded to the cloud.

[0034] The tower micro-deformation monitoring data is organized in time series form, with additional location identifiers and equipment status codes, to obtain a multi-dimensional monitoring data stream of the foundation structure response.

[0035] The multidimensional monitoring data stream and the modeling parameter set are spatiotemporally aligned and data interpolated to construct a joint observation matrix.

[0036] Based on the joint observation matrix, a Bayesian fusion framework is used to obtain the fused feature vector.

[0037] Based on the fused feature vector, structural equation modeling is used to analyze the direct influence path coefficients of each environmental factor on the deformation response, and to determine the main disaster-causing factors.

[0038] Based on the main disaster-causing factors, a spatiotemporal coupled feature dataset is generated and saved in Parquet columnar storage format.

[0039] Optionally, the spatiotemporal coupled feature dataset is input into the gradual risk calculus model to dynamically adjust the weights of each feature factor in risk assessment, specifically including: A gradual risk calculation model with a three-layer architecture consisting of a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer is constructed.

[0040] Based on the spatiotemporal coupled feature dataset, the correlation between each environmental factor and structural deformation and the current response sensitivity are calculated.

[0041] Based on the correlation and the current response sensitivity, the Q-learning algorithm is used to treat weight adjustment as an action space and risk prediction error as a reward function to update the weight allocation strategy.

[0042] Based on the aforementioned weight allocation strategy, weight priors are set according to the main disaster-causing mechanisms of different seasons.

[0043] Based on the aforementioned weight priors, a weight smoothing constraint is set.

[0044] Based on the weight smoothing constraint, a dynamically optimized risk weight vector is output.

[0045] Optionally, based on the risk calculation results after weight optimization, the risk progression trend is determined, and a warning level and judgment result are generated, specifically including: A weighted risk score is calculated based on the risk weight vector. A differential spatiotemporal risk evolution trajectory model is constructed based on the weighted risk score, and the risk progression trend is determined.

[0046] Based on the aforementioned risk progression trend, a three-level dynamic early warning threshold is set to determine the risk classification and obtain the early warning level.

[0047] Based on the aforementioned warning level, the DBSCAN algorithm is used to spatially cluster the risk scores of adjacent towers to identify high-risk contiguous areas.

[0048] Based on the warning level, a warning result report with timeliness is generated; the warning result report includes: warning level, expected instability time window, scope of impact, ranking of main control factors, and confidence score.

[0049] The warning results report is compared with the expert rule base. If there is a conflict, it is transferred to the manual review process.

[0050] Once the warning is confirmed to be effective, the warning level and the warning result report are sent to the operation and maintenance emergency response system.

[0051] Secondly, this application provides a landslide risk early warning analysis system for power transmission towers in the rainy season of plateau regions, which uses the aforementioned landslide risk early warning analysis method for power transmission towers in the rainy season of plateau regions for early warning analysis.

[0052] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method and system for early warning analysis of landslide risks on power transmission towers during the rainy season in plateau regions. By collecting multi-source environmental data, it extracts micro-topographic relief, soil plasticity index, effective pore water pressure, root activity index, freeze-thaw cycle frequency, and surface evapotranspiration ratio, and maps these data to a distribution feature space to construct a high-resolution, multi-dimensional environmental modeling parameter set. This allows for the accurate selection of core influencing indicators suitable for landslide disasters on power transmission towers during the rainy season in plateau regions, discarding invalid and redundant data. It acquires data on tower foundation displacement, foundation inclination angle, and dynamic stress changes, and fuses this data with the multi-dimensional environmental modeling parameter set to obtain a spatiotemporally coupled feature dataset. This achieves deep spatiotemporal coupling between external environmental inducing factors and tower deformation response data, thus providing a comprehensive analysis of the risks associated with power transmission tower landslides. It breaks through the limitations of traditional analysis relying solely on environmental data or single tower monitoring data; it inputs spatiotemporally coupled feature datasets into a gradual risk calculation model, dynamically adjusting the weights of each feature factor in risk assessment; it can adaptively match the actual impact of each core feature factor based on the evolution patterns of landslide disasters at different times and in different terrains during the rainy season on the plateau, solving the problems of traditional fixed-weight risk models being unable to adapt to the dynamic changes in the plateau environment and rigid risk assessment; based on the risk calculation results after weight optimization, it identifies the progressive trend of landslide risk and generates warning levels and assessment results; it can accurately classify warning levels and output clear judgment conclusions, solving the problems of traditional early warnings being unable to predict the progressive trend of landslide risk, warning lag, and ambiguous classification. This application addresses the challenges of complex terrain, variable weather, and the insidious and gradual evolution of landslide risks during the rainy season in high-altitude regions. Through multi-source data collection, standardized processing, refined modeling of core features, deep spatiotemporal data fusion, dynamic weighted risk calculation, and progressive risk assessment, a dynamic early warning system for power transmission tower landslides has been constructed, adaptable to the unique conditions of high-altitude areas. This system effectively solves the industry pain points of traditional tower landslide early warning systems, such as single-dimensional data, poor model adaptability, fixed and rigid weights, delayed risk prediction, and insufficient accuracy. It can accurately, dynamically, and in real-time perceive the evolution trend of landslide risks on power transmission towers during the rainy season in high-altitude regions, achieving tiered and precise early warning. This significantly improves the initiative, scientific rigor, and accuracy of landslide disaster risk prevention and control for high-altitude power transmission lines, effectively ensuring the safety of high-altitude power transmission tower equipment and the stability of power grid operation, and reducing the probability of power transmission safety accidents caused by landslides. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A flowchart illustrating a method for early warning analysis of landslide risks on power transmission towers during the rainy season in high-altitude areas, provided as an embodiment of this application; Figure 2 This is a schematic diagram of a process for collecting multi-source environmental data according to an embodiment of this application; Figure 3 A schematic diagram illustrating the process of constructing a high-resolution multidimensional environment modeling parameter set according to an embodiment of this application; Figure 4 This is a schematic diagram of a process for standardizing and preprocessing multi-source environmental data, provided as an embodiment of this application. Figure 5 This is a schematic diagram of the process for obtaining a spatiotemporal coupling feature dataset according to an embodiment of this application; Figure 6 A flowchart illustrating the dynamic adjustment of the weights of various feature factors in risk assessment, provided as an embodiment of this application; Figure 7 This is a schematic diagram illustrating the process of risk assessment and the generation of early warning levels and judgment results, provided in one embodiment of this application. Figure 8 A flowchart illustrating another method for early warning analysis of landslide risks on power transmission towers during the rainy season in high-altitude areas, provided as an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Example 1: In one exemplary embodiment, such as Figure 1As shown, a method for early warning analysis of landslide risks on power transmission towers during the rainy season in plateau regions is provided. This method includes the following steps: S1: Collect multi-source environmental data; the multi-source environmental data includes: environmental input signals and background information.

[0058] S2: Based on multi-source environmental data, extract core feature factors and map them to the distribution feature space to construct a high-resolution multi-dimensional environmental modeling parameter set; the core feature factors include: micro-topographic relief, soil plasticity index, effective pore water pressure, biological root activity index, freeze-thaw cycle frequency and surface evapotranspiration ratio.

[0059] S3: Acquire tower micro-deformation monitoring data and fuse it with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupling feature dataset; the micro-deformation monitoring data includes: foundation displacement, foundation tilt angle and dynamic stress change data.

[0060] S4: Input the spatiotemporal coupled feature dataset into the gradual risk calculus model and dynamically adjust the weight of each feature factor in risk determination; the gradual risk calculus model includes: factor sensitivity assessment layer, weight allocation layer and feedback adjustment layer.

[0061] S5: Based on the risk calculation results after weight optimization, determine the progressive trend of landslide risk and generate early warning level and judgment result.

[0062] Implementing steps S1 to S5 above enables the full-process, dynamic, and high-precision perception and graded early warning of landslide risks on plateau power transmission lines. This significantly improves the scientific, proactive, and accurate nature of landslide risk prevention and control for plateau power transmission lines, effectively avoids safety hazards such as tower damage and line outages caused by landslide disasters, and ensures the safe and stable operation of the plateau power grid.

[0063] As an optional implementation method, such as Figure 2 As shown, in step S1, multi-source environmental data is collected, specifically including: S11: Acquire multidimensional environmental input signals from several sensing systems; the multidimensional environmental input signals include: micro-topographic information, land physical indicators, surface hydrological changes, regional meteorological evolution, and biological indicator characteristics; the biological indicator characteristics include: changes in plant root density, evolution of surface moss coverage, and frequency of animal activity traces, used to reflect the long-term impact of biological activity on slope stability.

[0064] S12: Obtain macro-environmental background information from a remote big data source; the macro-environmental background information includes: regional meteorological, geological and vegetation distribution information, in order to obtain comprehensive and multi-dimensional environmental input signals.

[0065] S13: Perform time synchronization and spatial registration on the multidimensional environmental input signal and the macroscopic environmental background information to obtain synchronized and registered data.

[0066] S14: After classifying the synchronized and registered data, store it in a distributed time-series database.

[0067] S15: Perform data anomaly processing on the multidimensional environmental input signal and the macroscopic environmental background information to obtain anomaly-processed multi-source environmental data.

[0068] S16: Based on the multi-source environmental data after the anomaly processing, set the initial static weights.

[0069] This step breaks through the limitations of traditional monitoring that only focuses on single geological or deformation parameters. For the first time, it incorporates plateau-specific ecological-climate-geological composite factors such as biological root activity, surface evapotranspiration, and freeze-thaw cycles into a unified data collection framework, forming a raw data foundation with spatial heterogeneity characterization capabilities.

[0070] As an optional implementation method, such as Figure 3 As shown, in step S2, based on multi-source environmental data, core feature factors are extracted and mapped to the distribution feature space to construct a high-resolution multi-dimensional environmental modeling parameter set, specifically including: S21: Perform standardization preprocessing on multi-source environmental data to obtain standardized multi-source environmental data.

[0071] S22: Perform feature filtering on the standardized multi-source environmental data to obtain core feature factors.

[0072] S23: Extract the instantaneous values ​​of each characteristic factor; among them, the micro-topographic relief is calculated from laser point cloud data; the soil plasticity index is obtained by fitting the field electrical impedance measurement value to the laboratory calibration curve; the effective pore water pressure is derived from the Terzaghi effective stress principle; the biological root activity index is weighted by the root impedance change rate and the NDVI growth rate; the freeze-thaw cycle frequency is accumulated by the number of times the daily highest / lowest temperature crosses 0℃; the surface evapotranspiration ratio is determined by the ratio of the measured value of the evapotranspiration meter to the theoretical value of potential evapotranspiration.

[0073] S24: Based on the instantaneous values ​​of each feature factor, construct a multidimensional feature space to obtain the state distribution cloud map of the current monitoring area.

[0074] S25: Based on the state distribution cloud map, a time sliding window mechanism is introduced for each feature factor to extract dynamic evolution features.

[0075] S26: Based on the dynamic evolution characteristics, each feature factor is rasterized and assigned spatial weights to obtain a high-resolution multidimensional environment modeling parameter set.

[0076] S27: Encapsulate the multidimensional environment modeling parameter set into an extensible XML Schema structure.

[0077] This step extracts core environmental feature factors and constructs a high-resolution, multi-dimensional environmental modeling parameter set. Based on a standardized environmental dataset, core feature factors such as TRI, PI, EPWP, RAI, FTC, and ETR are extracted, and each factor is mapped to a distribution feature space to construct a high-resolution, multi-dimensional environmental modeling parameter set around the tower, characterizing the impact of differential microenvironments. This modeling parameter set not only includes static geographic attributes but also incorporates time-series feature representations, enabling a refined characterization of non-uniform environmental stress under typical plateau geomorphic units (such as slopes, valleys, and terraces).

[0078] As an optional implementation method, such as Figure 4 As shown, in step S21, the multi-source environmental data undergoes standardization preprocessing to obtain standardized multi-source environmental data, specifically including: S211: Standardize the physical quantity units and data format of the multi-source environmental data to obtain unified data.

[0079] S212: An adaptive filtering algorithm is used to remove high-frequency noise and drift interference from the unified data to obtain filtered data.

[0080] S213: Normalize the filtered data to obtain normalized data.

[0081] S214: Filter the normalized data based on the data quality scoring mechanism to obtain filtered data.

[0082] S215: Based on the filtered data, construct a version management mechanism.

[0083] S216: Based on the aforementioned management mechanism, output standardized multi-source environmental data.

[0084] This step performs environmental dataset standardization preprocessing, unifying the format, removing noise, and normalizing physical quantities of the collected multi-source environmental data to generate a standardized environmental dataset. This ensures that multidimensional variables in subsequent analyses have unified comparison and fusion conditions. This step employs adaptive filtering and a cross-modal Z-score normalization algorithm, effectively addressing data distortion issues caused by differences in sampling frequencies and dimensions of different sensors, as well as the low signal-to-noise ratio environment at high altitudes. This provides reliable data support for high-precision modeling.

[0085] As an optional implementation method, such as Figure 5As shown, in step S3, tower micro-deformation monitoring data is acquired and fused with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupled feature dataset, specifically including: S31: Based on the deployed multi-level micro-deformation sensing system, micro-deformation monitoring data of the transmission tower is acquired. The multi-level micro-deformation sensing system includes: a displacement sensor, a MEMS inclinometer, and a deep inclinometer. All sensing data are aggregated through an edge computing gateway, pre-compressed, and then uploaded to the cloud. The displacement sensor is a single fiber optic grating sensor, buried under the four legs of the transmission tower, with an accuracy of ±0.1mm. The MEMS inclinometer is installed on the side of the foundation, with a sampling frequency of 1Hz. The deep inclinometer is deployed along the potential sliding path to monitor deep soil displacement.

[0086] S32: Organize the tower micro-deformation monitoring data in time series form, add location identifiers and equipment status codes to obtain a multi-dimensional monitoring data stream of the foundation structure response.

[0087] S33: Perform spatiotemporal alignment and data interpolation matching between the multidimensional monitoring data stream and the modeling parameter set to construct a joint observation matrix.

[0088] S34: Based on the joint observation matrix, a Bayesian fusion framework is used to obtain the fused feature vector.

[0089] S35: Based on the fused feature vector, a structural equation model is used to analyze the direct influence path coefficients of each environmental factor on the deformation response, and to determine the main disaster-causing factors.

[0090] S36: Based on the main disaster-causing factors, generate a spatiotemporal coupled feature dataset and save it in Parquet columnar storage format.

[0091] This step involves monitoring and fusing micro-deformation data of the tower foundation. It utilizes surface and subsurface micro-deformation sensors of the tower foundation and surrounding soil in different regions to acquire data on foundation displacement, inclination angle, and dynamic stress changes at various locations. This data is then fused with a high-resolution environmental modeling parameter set to form a spatiotemporally coupled feature dataset. This step achieves bidirectional correlation modeling between external environmental driving factors and internal structural responses, enhancing the ability to identify the initiation and expansion processes of potential sliding surfaces.

[0092] As an optional implementation method, such as Figure 6 As shown, in step S4, the spatiotemporal coupled feature dataset is input into the gradual risk calculus model, and the weights of each feature factor in risk determination are dynamically adjusted, specifically including: S41: Construct a gradual risk calculus model with a three-layer architecture consisting of a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer.

[0093] S42: Based on the spatiotemporal coupled feature dataset, calculate the correlation between each environmental factor and structural deformation and the current response sensitivity.

[0094] S43: Based on the correlation and the current response sensitivity, the Q-learning algorithm is used to update the weight allocation strategy by treating weight adjustment as the action space and risk prediction error as the reward function.

[0095] S44: Based on the weight allocation strategy, set weight priors according to the main disaster-causing mechanisms of different seasons; S45: Based on the weight prior, set weight smoothing constraints.

[0096] S46: Based on the weight smoothing constraint, output the dynamically optimized risk weight vector.

[0097] This step constructs a gradual risk calculus model with a three-layer architecture: a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer. The spatiotemporally coupled feature dataset is input into the gradual risk calculus model, and a reinforcement learning algorithm is used to dynamically adjust the weights of each feature factor in risk assessment, achieving dynamic adaptive optimization of the gradual risk calculus model. This mechanism overcomes the rigidity of fixed-weight models, enabling the system to respond sensitively to seasonal changes, extreme weather events, or long-term degradation processes.

[0098] As an optional implementation method, such as Figure 7 As shown, in step S5, based on the risk calculation results after weight optimization, the risk progression trend is determined, and a warning level and judgment result are generated, specifically including: S51: Calculate the weighted risk score based on the risk weight vector, construct a differential spatiotemporal risk evolution trajectory model based on the weighted risk score, and determine the risk progression trend.

[0099] S52: Based on the aforementioned risk progression trend, a three-level dynamic early warning threshold is set to determine the risk classification and obtain the early warning level.

[0100] S53: Based on the aforementioned warning level, the DBSCAN algorithm is used to perform spatial clustering of the risk scores of adjacent towers to identify high-risk contiguous areas.

[0101] S54: Based on the warning level, generate a warning result report with timeliness; the warning result report includes: warning level, expected instability time window, scope of impact, ranking of main control factors and confidence score.

[0102] S55: Compare the aforementioned warning result report with the expert rule base. If there is a conflict, proceed to the manual review process.

[0103] S56: When the warning is confirmed to be effective, the warning level and the warning result report are sent to the operation and maintenance emergency response system.

[0104] This step identifies the progressive trend of landslide risk and generates an early warning level. Based on the risk calculation results optimized by weights, it identifies the gradual progressive trend of risk in the tower foundations of each sub-region. A differential spatiotemporal analysis method is used to determine whether the risk change exceeds a preset threshold. If an area is determined to be abnormally increasing, the corresponding landslide risk early warning level and timeliness determination result are automatically generated. This criterion considers not only the absolute risk value but also the rate of change and acceleration characteristics, improving the sensitivity of early hazard identification.

[0105] Example 2: like Figure 8 As shown, as another optional implementation, the method for early warning analysis of landslide risks on transmission towers during the rainy season in plateau regions may further include: S1: Collect multi-source environmental data; the multi-source environmental data includes: environmental input signals and background information.

[0106] S2: Perform standardization preprocessing on multi-source environmental data to obtain a standardized environmental dataset.

[0107] S3: Based on a standardized environmental dataset, extract core feature factors and map them to a distributed feature space to construct a high-resolution multidimensional environmental modeling parameter set; the core feature factors include: TRI, PI, EPWP, RAI, FTC and ETR.

[0108] S4: Acquire tower micro-deformation monitoring data and fuse it with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupling feature dataset; the micro-deformation monitoring data includes: foundation displacement, foundation tilt angle and dynamic stress change data.

[0109] S5: Input the spatiotemporal coupled feature dataset into the gradual risk calculus model and dynamically adjust the weight of each feature factor in risk determination; the gradual risk calculus model includes: factor sensitivity assessment layer, weight allocation layer and feedback adjustment layer.

[0110] S6: Based on the risk calculation results after weight optimization, determine the progressive trend of landslide risk and generate early warning level and judgment result.

[0111] S7: Send the warning level and judgment result to the operation and maintenance emergency response system, automatically match the response strategy library according to the warning level, and dynamically activate the emergency linkage response strategy; the response strategy includes: on-site inspection and dispatch, risk area isolation or temporary control instructions.

[0112] S8: Monitor the effectiveness of early warning implementation, record actual risk variations and environmental factor feedback data, and feed them back to the multidimensional environmental modeling parameter set, periodically executing the model self-updating algorithm.

[0113] By implementing steps S1 to S8 above, and constructing an integrated intelligent analysis chain of "perception-modeling-fusion-decision-feedback," a fundamental shift from passive monitoring to proactive early warning is achieved. Through the synergistic effect of a dynamic weight optimization mechanism and a self-updating model structure, the system can capture subtle trend changes caused by accumulated environmental stresses before significant deformation occurs, thus significantly advancing the early warning window. Simultaneously, multi-dimensional factor joint modeling significantly enhances the ability to suppress false alarms under complex geological backgrounds, improving the reliability of early warning results.

[0114] This significantly improves the comprehensiveness, timeliness, and accuracy of landslide risk identification for transmission towers in high-altitude areas, effectively reducing the risk of power outages caused by geological disasters, greatly enhancing the safety assurance level and emergency response efficiency of power grid operation and maintenance, and fully ensuring the stable operation of power infrastructure in high-altitude areas. Through multi-source data fusion and adaptive modeling, it effectively overcomes the shortcomings of traditional methods in adapting to complex environments, reliably enhances the intelligence level of risk assessment, and significantly improves the practicality and maintainability of the early warning system.

[0115] As an optional implementation, in step S7, disaster risk priority early warning and dynamic operation and maintenance response instructions are pushed out. The early warning level and judgment results are output to the operation and maintenance emergency response system in real time. Priority information is automatically generated for different risk levels, and disaster emergency response strategies are dynamically activated, including instructions such as on-site inspection and dispatch, risk area isolation, or temporary control, to achieve proactive early warning and dynamic response to landslide risks in plateau tower areas. This response mechanism has the functions of hierarchical triggering, resource matching, and path optimization, which significantly improves the efficiency of emergency response.

[0116] In step S7, the warning levels include: yellow warning, orange warning, and red warning.

[0117] When the warning level is yellow, a command to dispatch drones for patrol is generated.

[0118] When the warning level is orange, an order is generated to dispatch an on-site investigation team.

[0119] When the warning level is red, an instruction is generated to activate the power outage plan and seal off the danger zone.

[0120] As an optional implementation, in step S8, a feedback and model self-updating mechanism is implemented to monitor the effectiveness of the early warning system. The feedback data on actual risk variations and changes in environmental factors are recorded and analyzed. The feedback results are input into a multi-dimensional environmental modeling parameter set, and the model self-updating algorithm is periodically executed to optimize the accuracy and timeliness of subsequent risk assessments, thereby achieving continuous iterative improvement in system performance. This closed-loop mechanism endows the system with self-evolution capabilities, ensuring high robustness and predictive reliability during long-term operation.

[0121] Example 3: In practical applications, the method for early warning analysis of landslide risks on transmission towers in plateau rainy seasons includes the following steps: S1: Collect multi-source environmental data. Simultaneously acquire multi-dimensional environmental data such as micro-topography information, soil physical indicators, surface hydrological changes, regional meteorological evolution, and biological indicators from multiple environmental monitoring collection points deployed around the high-altitude power transmission towers. Integrate these data with regional meteorological forecasts, geological maps, and vegetation cover remote sensing images from a remote big data platform to form a set of original environmental input signals with clear spatial distribution and continuous temporal sequence.

[0122] S2: Perform environmental data standardization preprocessing. For the heterogeneous multi-source environmental data obtained in step S1, adopt a unified data format conversion protocol, remove abnormal noise values ​​and implement cross-modal normalization processing to generate a standardized environmental dataset with dimensional consistency and comparability, providing high-quality data support for subsequent feature extraction and modeling.

[0123] S3: Extract environmental feature factors and construct high-resolution modeling parameters. Based on the standardized dataset output in step S2, extract the core environmental feature factors that reflect the typical disaster-causing mechanisms of the plateau, map them to a multi-dimensional feature space, and construct a high-resolution modeling parameter set that characterizes the differential micro-environmental stress around the tower.

[0124] S4: Perform micro-deformation monitoring and coupled data fusion of the tower foundation. The surface and underground micro-deformation sensors deployed in the tower foundation and its affected area are called by region to collect structural behavior data such as foundation displacement, tilt angle change and stress response. The data are then spatiotemporally aligned and deeply fused with the modeling parameter set obtained in step S3 to generate a coupled feature dataset containing both environmental driving and structural response dimensions.

[0125] S5: Construct a gradual risk calculus model with a three-layer architecture: a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer. Input the aforementioned spatiotemporally coupled feature dataset into the gradual risk calculus model, and dynamically adjust the weights of each feature factor in risk determination using a reinforcement learning algorithm. This achieves dynamic adaptive optimization of the gradual risk calculus model, enabling precise adaptation to different stages of disaster evolution. The gradual risk calculus model, through its three-layer architecture (sensitivity assessment, weight allocation, and feedback adjustment), dynamically assesses factors by combining historical correlation and immediate sensitivity, and employs a reinforcement learning mechanism incorporating seasonal priors for adaptive weight optimization. Simultaneously, smoothing constraints are applied to ensure weight stability. S6: Determine the progressive trend of landslide risk and generate early warning levels. Based on the weighted risk calculation results output in step S5, use differential spatiotemporal analysis technology to identify the rate of change and acceleration characteristics of risk values ​​in each zone, and determine whether there is a significant increasing trend. If the threshold is exceeded, it is determined to be a potential high-risk area for landslides, and an early warning signal and timeliness assessment of the corresponding level are automatically generated.

[0126] S7: Push disaster risk priority warning and dynamic operation and maintenance response instructions. Push the warning level and location information generated in step S6 to the power grid operation and maintenance emergency response system in real time. Automatically trigger graded linkage strategies based on risk level, including inspection task assignment, traffic control suggestions or temporary power outage plans, to improve disaster response efficiency.

[0127] S8: Implement a feedback and model self-update mechanism to continuously monitor the actual effect of the early warning response in step S7, collect post-verification data (such as field survey reports, GNSS re-measurement results) and new environmental observation streams, feed them back to the modeling system, periodically execute model parameter correction and structural optimization algorithms, and complete the closed-loop self-update of the gradual risk calculation model.

[0128] Step S1: Collect multi-source environmental data, specifically including: S1.1: Construct a multi-scale environmental monitoring network around high-altitude power transmission towers.

[0129] This sub-step aims to establish a three-dimensional sensing system covering the core area of ​​the transmission tower and its extended influence range. Several fixed environmental monitoring nodes are deployed within a radius of 50-200 meters from the tower foundation. Each node integrates a lidar module to collect micro-topographical undulation data, a soil triaxial sensor array to measure moisture content, density, and shear strength, and a micro-weather station to record temperature, precipitation, wind speed, and solar radiation. Simultaneously, a surface evapotranspiration meter and a freeze-thaw cycle monitoring probe are installed to capture the unique thermodynamic boundary processes of the plateau. Fiber optic displacement sensors with an accuracy of ±0.1mm are buried under the four legs of the transmission tower; MEMS inclinometers with a sampling frequency of 1Hz are installed on the sides of the foundation; and deep inclinometers are deployed along the potential sliding path to monitor deep soil displacement. Furthermore, root activity monitoring strips are deployed in densely vegetated areas to indirectly assess changes in the soil-fixing capacity of plant roots using the micro-current impedance method. All sensor data is aggregated through an edge computing gateway, pre-compressed, and uploaded to the cloud. All local sensors are networked via LoRa wireless communication to ensure stable long-distance transmission under low-power conditions.

[0130] S1.2: Obtain macro-environmental background information from a remote big data source.

[0131] This sub-step focuses on introducing large-scale environmental background data to enhance the contextual understanding capabilities of local monitoring. Medium- and long-term meteorological trends, regional geological structure maps, seismic activity records, and NDVI vegetation index time-series layers for the target area are periodically acquired from the National Meteorological Administration, the Ministry of Natural Resources' Geological Survey Information System, and the MODIS remote sensing database via API interfaces. After downloading the data in GeoTIFF or NetCDF format, it is unified to the WGS84 geographic coordinate system through coordinate projection correction and matched to the local monitoring grid at a daily granularity, thereby achieving spatial-temporal alignment between micro-level field data and macro-level background information.

[0132] S1.3: Implement time synchronization and spatial registration of multi-source heterogeneous data.

[0133] Because various sensors have different sampling frequencies (ranging from seconds to days), strict timestamp alignment is required. A GPS timing module provides a UTC standard time reference for all local devices, while remote data is linearly interpolated to fill in missing time periods based on its release time. Spatially, a GIS platform is used to perform inverse distance-weighted interpolation (IDW) of point sensor readings into a 2m×2m resolution raster surface, allowing data from different sources to be expressed within the same geographic reference frame, forming a multidimensional environmental field with spatial continuity.

[0134] S1.4: Categorize and store the original environmental input signals and establish metadata indexes.

[0135] All collected data is categorized and stored in a distributed time-series database (such as InfluxDB) according to a three-level directory structure of "device type-geographic location-timestamp". Each record is accompanied by complete metadata tags, including sensor model, installation depth, calibration date, and effective range. This design supports efficient querying and version tracking, facilitating later quality control and model training.

[0136] S1.5: Design an automatic alarm and retransmission mechanism for abnormal data acquisition events.

[0137] When a node fails to upload data three times consecutively or the data value exceeds the historical fluctuation range of ±3σ, the system automatically marks it as "data abnormal" and sends an alarm notification to the operation and maintenance center via the backup satellite link. At the same time, the neighboring node data compensation mechanism is activated, using Kriging interpolation to estimate the missing measurement value, and performing data retransmission verification after communication is restored to ensure data integrity.

[0138] S1.6: Define the initial weight configuration scheme for multidimensional environmental factors.

[0139] To lay the foundation for subsequent dynamic optimization, initial static weights were set based on expert experience and existing literature: micro-topography 20%, soil mechanical properties 25%, hydrological changes 20%, meteorological evolution 15%, biological indicators 10%, and remote background information 10%. This configuration serves as the starting point for adaptive learning in S5 and can be gradually adjusted as the system operates.

[0140] Step S2: Perform environmental data standardization preprocessing, specifically including: S2.1: Standardize the physical quantity units and data format specifications for multi-source data.

[0141] Sensor outputs from different manufacturers were uniformly converted to the International System of Units (SI): displacement was expressed in millimeters (mm), temperature in degrees Celsius (°C), pressure in kilopascals (kPa), and conductivity in mS / cm. Textual information (such as geological lithology descriptions) was encoded as One-Hot vectors, and image-based remote sensing data was converted to 8-bit grayscale or RGB standardized format. All data was ultimately encapsulated in a JSON-LD structure with embedded semantic tags for machine parsing.

[0142] S2.2: Apply an adaptive filtering algorithm to remove high-frequency noise and drift interference.

[0143] To address sensor drift caused by strong winds and large diurnal temperature variations at high altitudes, an improved wavelet thresholding denoising algorithm (Daubechies db4 basis function) is employed to decompose and reconstruct the time series signal. For low-frequency trend terms, Hodrick-Prescott filtering is used to separate long-term variations from short-term disturbances; for sudden spike noise, mean filtering combined with Grubbs' test is applied to identify and remove outliers.

[0144] S2.3: Implement cross-modal normalization to eliminate dimensional differences.

[0145] Considering the vast differences in the numerical ranges of various environmental factors (e.g., precipitation ranges from 0 to 50 mm / d, while topographic elevation can reach several kilometers), the Z-score normalization formula is adopted: ; in, Here, x represents the original value, u is the mean of the variable over the past 30 days, and σ is its standard deviation. For non-normally distributed variables (such as rainfall), use Min-Max normalization instead. ; in, The corresponding minimum value; This represents the corresponding maximum value.

[0146] Ensure that all variables fall within the range of [-1, 1] or [0, 1] to avoid certain factors dominating the modeling process due to excessively large values.

[0147] S2.4: Establish a data quality scoring mechanism and filter low-confidence samples.

[0148] Each data record is assigned a quality score Q (0 ≤ Q ≤ 1), calculated by comprehensively considering factors such as signal stability, equipment health status, and environmental interference. When Q < 0.6, the data is considered unreliable and will not be included in subsequent modeling. The scoring model uses a fuzzy logic rule engine, for example: "If the coefficient of variation of 5 consecutive sampling points is > 50%, and there is no precipitation record during the same period, then the score is reduced by 0.2."

[0149] S2.5: Establish a version management mechanism for standardized environment datasets.

[0150] After each preprocessing step, a new data version package is generated, named "ENV_STD_vYYYYMMDD_HHMM", and processing parameters (such as filter window size and normalization method selection) are recorded. Older versions are retained for at least 90 days to support model rollback and comparative experiments.

[0151] S2.6: Output a standardized environment dataset for downstream modules to use.

[0152] The final standardized dataset is stored in HDF5 format, supporting fast random access and compressed transmission. It is provided externally via a RESTful API, allowing step S3 to retrieve data slices of specific regions, time periods, and factor combinations on demand.

[0153] Step S3: Extract environmental feature factors and construct high-resolution modeling parameters, specifically including: S3.1: Define the set of key disaster-causing characteristic factors for high-altitude power transmission tower areas.

[0154] Based on research on the mechanisms of geological hazards in plateau regions, six core characteristic factors were identified: TRI, PI, EPWP, RAI, FTC, and ETR. These factors represent key disaster-causing pathways such as topographic driving, material degradation, water effects, ecological reinforcement, thermal disturbance, and climate forcing.

[0155] S3.2: Quantitatively extract the instantaneous values ​​of each feature factor from standardized data.

[0156] TRI is calculated from laser point cloud data, and the formula is: ; in, For the elevation of the central grid point, The elevation of the surrounding 8 neighboring points.

[0157] PI is obtained by fitting the field impedance measurement value to the laboratory calibration curve.

[0158] EPWP is derived using the Terzaghi effective stress principle: in, The depth of the groundwater level. Here, denoted by , is the specific weight of water, and u represents the excess pore water pressure (EPWP).

[0159] RAI is a weighted synthesis of the root impedance change rate and the NDVI growth rate.

[0160] FTC is calculated by the number of times the daily maximum / minimum temperature crosses 0°C.

[0161] ETR is determined by the ratio of the measured value of the evapotranspiration meter to the theoretical value of the potential evapotranspiration.

[0162] S3.3: Construct a multidimensional feature space and implement factor mapping.

[0163] Using the six characteristic factors mentioned above as coordinate axes, a six-dimensional feature space is constructed, with each spatial location corresponding to a typical microenvironmental state. The kernel density estimation (KDE) method is used to depict the state distribution cloud map of the current monitoring area in this space, identifying cluster centers as typical environmental pattern prototypes.

[0164] S3.4: Introduce a time-sliding window mechanism to extract dynamic evolution features.

[0165] For each feature factor, a 7-day sliding window is constructed, and its mean, variance, slope, and skewness are calculated to form a composite feature vector of "static + dynamic". For example, EPWP not only records the current value, but also includes its growth trend (regression slope) and the degree of fluctuation (standard deviation) over the past week.

[0166] S3.5: Generate a high-resolution modeling parameter set and assign spatial weights.

[0167] The feature factors are rasterized into a 2m resolution map layer. Combined with the projected area of ​​the tower foundation, the influence weight of each grid cell on the tower stability is calculated. The formula is: ; in, For Euler number, For grid Distance to the center of the tower m is the attenuation constant. The final modeling parameter set is a weighted feature matrix, used to characterize the spatial variability of non-uniform environmental stress.

[0168] S3.6: Encapsulate the modeling parameter set into an extensible XML Schema structure.

[0169] Define a unified data structure schema, including fields: <featurename> 、 <value> 、 <uncertainty> 、 <timestamp> 、 <spatialweight>This allows for seamless integration of newly added feature factors in the future, improving system maintainability.

[0170] Step S4: Perform micro-deformation monitoring and coupled data fusion of the tower foundation, specifically including: S4.1: Deploy a multi-layered micro-deformation sensing system.

[0171] Fiber Bragg grating (FBG) displacement sensors with an accuracy of ±0.1 mm are embedded under the four legs of the tower; MEMS inclinometers with a sampling frequency of 1 Hz are installed on the sides of the foundation; and deep inclinometers are deployed along the potential sliding path to monitor deep soil displacement. All sensor data are aggregated through an edge computing gateway, pre-compressed, and then uploaded to the cloud.

[0172] S4.2: Obtain multi-dimensional monitoring data stream of infrastructure response.

[0173] The data collected in real time includes: vertical settlement. Horizontal displacement Twist angle Foundation bottom stress Lateral earth pressure The data stream is organized in a time-series format, with additional location identifiers and device status codes.

[0174] S4.3: Perform spatiotemporal alignment and data interpolation matching.

[0175] Because environmental data has a long update cycle (e.g., once a day), while deformation data is on a minute-by-minute basis, the former needs to be upsampled to the same time granularity. Cubic spline interpolation is used to fill in intermediate values ​​of environmental parameters, and these are precisely aligned with the deformation data by timestamp to construct a joint observation matrix. ,in, For environment vectors, This is the deformation vector.

[0176] S4.4: Apply a Bayesian fusion framework to integrate environmental and structural data.

[0177] Construct a probabilistic graphical model, assuming environmental factors are latent variables. The observed deformation is the manifest variable. Then the posterior probability is: ; Among them, prior Likelihood provided by modeling in step S3 It was learned from historical training data. X is a latent variable, and X is a manifest variable. The most likely environment-structure coupling state is inferred through maximum a posteriori estimation (MAP), generating a fused feature vector. , The most likely environmental factor value is selected.

[0178] S4.5: Introduce a causal reasoning mechanism to identify the main disaster-causing path.

[0179] Structural equation modeling (SEM) was used to analyze the direct influence path coefficients of various environmental factors on the deformation response, identifying the main causative factors. For example, if EPWP→ If the path coefficient is significantly higher than other paths, it is determined that the current stage is mainly controlled by hydrological factors.

[0180] S4.6: Outputs a spatiotemporal coupled feature dataset for use by the risk calculation model.

[0181] The final dataset is stored in Parquet columnar storage format, containing fields such as timestamp, spatial location, environmental features, deformation response, and fusion confidence, supporting large-scale parallel processing and machine learning modeling.

[0182] Step S5: Perform dynamic optimization analysis of risk gradient factor weights, specifically including: S5.1: Construct a gradual risk calculation model with a three-layer architecture: factor sensitivity assessment layer, weight allocation layer, and feedback adjustment layer.

[0183] The model consists of three layers: a bottom layer for factor sensitivity assessment, a middle layer for weight allocation, and a top layer for feedback adjustment. The input is the fused feature dataset output from step S4, and the output is the risk contribution weight vector of each factor at the current time step. .

[0184] S5.2: Calculate the historical correlation and current response sensitivity of each factor.

[0185] The Pearson correlation coefficient between each environmental factor and structural deformation was calculated using a rolling window. The predictive power was assessed using Granger causality tests. The rate of change of factors within the current time period was also calculated. The resulting deformation response increment The ratio is used as an indicator of instantaneous sensitivity.

[0186] S5.3: Update weight allocation strategy based on reinforcement learning mechanism.

[0187] The Q-learning algorithm is employed, treating weight adjustment as the action space and risk prediction error as the reward function. (State) Actions are defined by the current environment composition. The weights are fine-tuned to minimize the long-term prediction loss. ; in, Mean square error, This is a discount factor. After multiple rounds of training, the model learns to automatically increase the weights of key factors under different environmental conditions.

[0188] S5.4: Introduce seasonal prior knowledge to guide weight initialization.

[0189] The prior weights are set according to the main disaster-causing mechanisms of different seasons: hydrological factors are emphasized in spring (weight +15%), freeze-thaw factors in winter (weight +20%), and biological factors in summer (weight +10%). This prior serves as the starting strategy for reinforcement learning, accelerating the convergence process.

[0190] S5.5: Set weighted smoothing constraints to prevent severe oscillations.

[0191] To avoid misjudgments caused by sudden weight changes, an L2 regularization term is introduced to limit the weight difference between adjacent time steps: in, , Contribute weight vectors to two adjacent risks. The threshold value is used. If the adjustment range is too large, an exponential moving average method is used for smoothing the transition.

[0192] S5.6: Output the dynamically optimized risk weight vector for subsequent discrimination.

[0193] The final weight vector is multiplied by the original feature vector to obtain the weighted risk score R: w k ·f k ; Among them, w k f is the generated k-th weight vector; k This is the k-th original feature vector.

[0194] This score serves as the core basis for trend identification in S6.

[0195] Step S6: Determine the progressive trend of landslide risk and generate an early warning level, specifically including: S6.1: Construct a differential spatiotemporal risk evolution trajectory model.

[0196] Weighted risk score Treating it as a function of time, calculate its first derivative (rate of change) and second derivative (acceleration): ; when and When this occurs, it indicates that the risk is accelerating and the area is entering a key concern zone.

[0197] S6.2: Set multi-level dynamic thresholds for risk classification and determination.

[0198] Define three warning thresholds: Yellow warning ( and / day), Orange Alert ( and / day 2 Red alert ( (Or sudden jumps may occur). The threshold can be dynamically adjusted based on the historical disaster frequency of the region.

[0199] S6.3: Introduce spatial clustering analysis to identify high-risk contiguous areas.

[0200] The DBSCAN algorithm is used to spatially cluster the risk values ​​of adjacent towers. If three or more towers enter the yellow warning zone consecutively, the overall warning level of the area is upgraded to prevent local misjudgments from affecting the overall judgment.

[0201] s6.4: Generates a timely warning result report.

[0202] The early warning report includes: risk level, expected instability time window (estimated based on extrapolation), scope of impact, ranking of main control factors, confidence score, etc., and is output in both PDF and JSON formats.

[0203] S6.5: Implement a multi-verification mechanism for early warning results.

[0204] The automatic judgment results are compared with the expert rule base. If there is a conflict (such as the model reporting red but there is no obvious deformation), the process is transferred to manual review to avoid misjudgment by a single algorithm.

[0205] S6.6: Trigger the early warning signal and transmit the decision data to step S7.

[0206] Once the warning is confirmed to be valid, the alarm channel is immediately activated, and the warning packet is pushed to the emergency response system via the MQTT protocol to initiate the next response procedure.

[0207] Step S7: Push disaster risk priority early warning and dynamic operation and maintenance response instructions, specifically including: S7.1: Establish an operation and maintenance resource scheduling model and path optimization algorithm.

[0208] Based on GIS network analysis, an emergency resource layer including road conditions, personnel locations, and material warehouses is constructed. The Dijkstra algorithm is used to solve for the optimal arrival path and generate inspection route planning.

[0209] S7.2: Automatically match the response strategy library according to the warning level.

[0210] Yellow alert: Dispatch drones for patrol; Orange alert: Organize on-site investigation teams; Red alert: Activate power outage contingency plan and seal off dangerous areas. Each strategy includes details such as personnel allocation, equipment list, and communication plan.

[0211] S7.3: Dynamically generate work orders and push them to mobile operation and maintenance terminals.

[0212] Electronic work orders can be issued via WeChat or a dedicated app, including task details, safety instructions, navigation links, and reporting deadlines, and support online receipt and progress feedback.

[0213] S7.4: Implement risk area isolation and temporary control measures.

[0214] The video surveillance system is linked to direct cameras toward high-risk areas, activates audio-visual warning devices to alert passersby, and broadcasts evacuation notices via the public address system.

[0215] S7.5: Record the entire response process log for later auditing and model optimization.

[0216] All operation times, responsible persons, and execution results are recorded in the blockchain evidence storage system to ensure traceability and tamper-proofness.

[0217] S7.6: Enables multi-departmental collaborative interface integration.

[0218] Establish data sharing channels with local government emergency management bureaus and traffic management departments to automatically report relevant information during major warnings and seek external support.

[0219] Step S8: Implement a feedback and model self-updating mechanism, specifically including: S8.1: Establish an indicator system for evaluating the effectiveness of early warning.

[0220] We define three core metrics: Precision, Recall, and Lead Time, and evaluate the effectiveness of each alert based on the results of on-site verification.

[0221] S8.2: Collect feedback data of actual events.

[0222] If a landslide occurs, collect real-world consequences data such as GNSS re-measurement data, crack width measurements, and photogrammetric 3D reconstruction models, and add them to the training set as negative samples.

[0223] S8.3: Analyze the causes of false alarms and missed alarms and locate model defects.

[0224] Analyze the causes of errors to determine whether they are due to data quality issues, missing features, or weight bias, and propose targeted improvement directions.

[0225] S8.4: Perform periodic model retraining and parameter fine-tuning.

[0226] The entire model is automatically retrained once a month, using incremental learning to update the neural network weights and maintain the model's timeliness.

[0227] S8.5: Update the prior knowledge base in the high-resolution modeling parameter set.

[0228] Newly discovered important disaster-causing patterns (such as precursors to vegetation degradation) are encoded into new feature rules, enriching the factor extraction logic in step S3.

[0229] S8.6: Complete the model version upgrade and implement A / B testing to verify the performance improvement.

[0230] Before the new model goes live, it is run in parallel with the old version for a period of time to compare the consistency and accuracy of the predictions. Only after confirming that there are no errors can a full switch be made.

[0231] Example 4: This application also provides a landslide risk early warning analysis system for power transmission towers in the rainy season of plateau regions, which uses the aforementioned landslide risk early warning analysis method for power transmission towers in the rainy season of plateau regions for early warning analysis.

[0232] Example 5: In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 9 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores multi-source environmental data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for early warning analysis of landslide risks on power transmission towers during the rainy season in high-altitude areas.

[0233] Figure 9 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0234] Example 6: In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0235] Example 7: In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0236] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.

[0237] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0238] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0239] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0240] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.< / spatialweight> < / timestamp> < / uncertainty> < / value> < / featurename>

Claims

1. A method for early warning and analysis of landslide risk on power transmission towers during the rainy season in plateau regions, characterized in that, include: Collect multi-source environmental data; Multi-source environmental data includes: environmental input signals and background information; Based on multi-source environmental data, core feature factors are extracted and mapped to the distribution feature space to construct a high-resolution multi-dimensional environmental modeling parameter set. The core feature factors include: micro-topographic relief, soil plasticity index, effective pore water pressure, biological root activity index, freeze-thaw cycle frequency and surface evapotranspiration ratio. Micro-deformation monitoring data of the tower is acquired and fused with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupling feature dataset. The micro-deformation monitoring data includes: foundation displacement, foundation tilt angle and dynamic stress change data. The spatiotemporal coupled feature dataset is input into the gradual risk calculus model to dynamically adjust the weights of each feature factor in risk assessment. The gradual risk calculus model includes: a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer. Based on the risk calculation results after weight optimization, the progressive trend of landslide risk is determined, and the warning level and judgment result are generated.

2. The method for early warning and analysis of landslide risk of transmission towers in plateau rainy season as described in claim 1, characterized in that, The method further includes: The warning level and judgment result are sent to the operation and maintenance emergency response system. The system automatically matches the response strategy library according to the warning level and dynamically activates the emergency response strategy. The emergency response strategy includes: on-site inspection and dispatch, risk area isolation or temporary control instructions. Based on the aforementioned emergency response strategy, actual risk variations and environmental factor feedback data are recorded and fed back to the multidimensional environmental modeling parameter set, and the model self-updating algorithm is executed periodically.

3. The method for early warning analysis of landslide risk of transmission towers in plateau rainy season as described in claim 2, is characterized in that, The warning levels include: yellow warning, orange warning, and red warning; When the warning level is yellow, a command to dispatch drones for patrol is generated; When the warning level is orange, an order is generated to dispatch an on-site investigation team. When the warning level is red, an instruction is generated to activate the power outage plan and seal off the danger zone.

4. The method for early warning analysis of landslide risk of transmission towers in plateau rainy season as described in claim 1, characterized in that, Collect multi-source environmental data, specifically including: Acquire multidimensional environmental input signals from several sensing systems; the multidimensional environmental input signals include: micro-topographic information, land physical indicators, surface hydrological changes, regional meteorological evolution, and biological indicator characteristics; the biological indicator characteristics include: changes in plant root density, evolution of surface moss coverage, and frequency of animal activity traces. Obtain macro-environmental background information from remote big data sources; the macro-environmental background information includes: regional meteorological, geological, and vegetation distribution information. The multidimensional environmental input signal and the macroscopic environmental background information are synchronized in time and registered in space to obtain synchronized and registered data. The synchronized and registered data are then categorized and stored in a distributed time-series database; Data anomaly processing is performed on the multidimensional environmental input signal and the macroscopic environmental background information to obtain anomaly-processed multi-source environmental data. Based on the multi-source environmental data after anomaly processing, initial static weights are set.

5. The method for early warning and analysis of landslide risk of transmission towers in plateau rainy season as described in claim 1, characterized in that, Based on multi-source environmental data, core feature factors are extracted and mapped to a distribution feature space to construct a high-resolution, multi-dimensional environmental modeling parameter set, specifically including: Standardized multi-source environmental data is obtained by performing standardized preprocessing on multi-source environmental data. Feature filtering is performed on the standardized multi-source environmental data to obtain core feature factors; Instantaneous values ​​of each characteristic factor were extracted; among them, the micro-topographic relief was calculated from laser point cloud data; the soil plasticity index was obtained by fitting the field electrical impedance measurement value to the laboratory calibration curve; the effective pore water pressure was derived from the Terzaghi effective stress principle; the biological root activity index was synthesized by weighted average of the root impedance change rate and the NDVI growth rate; the freeze-thaw cycle frequency was accumulated from the number of times the daily highest / lowest temperature crossed 0℃; the surface evapotranspiration ratio was determined by the ratio of the measured value of the evapotranspiration meter to the theoretical value of potential evapotranspiration. Based on the instantaneous values ​​of each feature factor, a multidimensional feature space is constructed to obtain the state distribution cloud map of the current monitoring area; Based on the state distribution cloud map, a time sliding window mechanism is introduced for each feature factor to extract dynamic evolution features; Based on the dynamic evolution characteristics, each feature factor is rasterized and assigned a spatial weight to obtain a high-resolution multidimensional environment modeling parameter set. The multidimensional environment modeling parameter set is encapsulated into an extensible XML Schema structure.

6. The method for early warning analysis of landslide risk of transmission towers in plateau rainy season as described in claim 5, is characterized in that, Standardized multi-source environmental data is obtained by performing standardization preprocessing on multi-source environmental data, specifically including: The physical quantity units and data format of the multi-source environmental data are standardized to obtain unified data. An adaptive filtering algorithm is used to remove high-frequency noise and drift interference from the unified data to obtain filtered data. The filtered data is then normalized to obtain normalized data. The normalized data is filtered based on a data quality scoring mechanism to obtain filtered data. Based on the filtered data, a version management mechanism is constructed; Based on the aforementioned management mechanism, standardized multi-source environmental data is output.

7. The method for early warning analysis of landslide risk of transmission towers in plateau rainy season as described in claim 1, characterized in that, Data on micro-deformation monitoring of the tower was acquired and fused with a multi-dimensional environmental modeling parameter set to obtain a spatiotemporal coupled feature dataset, specifically including: Based on the deployed multi-level micro-deformation sensing system, tower micro-deformation monitoring data are acquired; the multi-level micro-deformation sensing system includes: displacement sensor, MEMS inclinometer and deep inclinometer; all sensing data are aggregated through edge computing gateway, pre-compressed and then uploaded to the cloud; The tower micro-deformation monitoring data is organized in time series form, with additional location identifiers and equipment status codes, to obtain a multi-dimensional monitoring data stream of the foundation structure response. The multidimensional monitoring data stream and the modeling parameter set are spatiotemporally aligned and data interpolated to construct a joint observation matrix. Based on the joint observation matrix, a Bayesian fusion framework is used to obtain the fused feature vector; Based on the fused feature vector, structural equation modeling is used to analyze the direct influence path coefficients of each environmental factor on the deformation response, and to determine the main disaster-causing factors. Based on the main disaster-causing factors, a spatiotemporal coupled feature dataset is generated and saved in Parquet columnar storage format.

8. The method for early warning and analysis of landslide risk of transmission towers in plateau rainy season as described in claim 1, characterized in that, The spatiotemporally coupled feature dataset is input into the gradual risk calculus model, and the weights of each feature factor in risk assessment are dynamically adjusted, specifically including: A gradual risk calculus model with a three-layer architecture consisting of a factor sensitivity assessment layer, a weight allocation layer, and a feedback adjustment layer is constructed. Based on the spatiotemporal coupled feature dataset, the correlation between each environmental factor and structural deformation and the current response sensitivity are calculated. Based on the correlation and the current response sensitivity, the Q-learning algorithm is used to update the weight allocation strategy by treating weight adjustment as the action space and risk prediction error as the reward function. Based on the aforementioned weight allocation strategy, weight priors are set according to the main disaster-causing mechanisms of different seasons; Based on the aforementioned weight priors, weight smoothing constraints are set; Based on the weight smoothing constraint, a dynamically optimized risk weight vector is output.

9. The method for early warning analysis of landslide risk of transmission towers in plateau rainy season according to claim 1, characterized in that, Based on the risk calculation results after weight optimization, the risk progression trend is determined, and warning levels and judgment results are generated, specifically including: A weighted risk score is calculated based on the risk weight vector. A differential spatiotemporal risk evolution trajectory model is constructed based on the weighted risk score, and the risk progression trend is determined. Based on the aforementioned risk progression trend, a three-level dynamic early warning threshold is set to determine the risk classification and obtain the early warning level. Based on the aforementioned warning level, the DBSCAN algorithm is used to perform spatial clustering of the risk scores of adjacent towers to identify high-risk contiguous areas. Based on the warning level, a timely warning result report is generated; the warning result report includes: warning level, expected instability time window, scope of impact, ranking of main control factors, and confidence score; The warning results report is compared with the expert rule base. If there is a conflict, it is transferred to the manual review process. Once the warning is confirmed to be effective, the warning level and the warning result report are sent to the operation and maintenance emergency response system.

10. A landslide risk early warning and analysis system for power transmission towers in high-altitude rainy seasons, characterized in that, The landslide risk early warning analysis method for transmission towers in plateau rainy season as described in any one of claims 1-9 is used for early warning analysis.

Images