Power transmission disaster prevention monitoring system for industrial data management

Abstract

The present application relates to power transmission disaster prevention monitoring and industrial data management technical field, specifically to a kind of industrial data management's power transmission disaster prevention monitoring system, including multiple edge perception nodes distributed in target area, with edge perception node communication connection and having adjustable transmission bandwidth edge computing gateway and data processing hub;Edge perception node acquires industrial state data and node operating state data;Data processing hub receives data, quantization is calculated network fragile entropy state value of perception network based on node operating state data, change trend is calculated in combination with historical network fragile entropy state value sequence, predict the countdown rate of perception network approaching preset failure time threshold, and at least including regular polling instruction and suboptimal degradation scheduling instruction Dynamic scheduling instruction is generated and issued to edge perception node, to carry out integrated management and control to data transmission and computing power state.

Description

Technical Field

[0001] This invention relates to the field of power transmission disaster prevention monitoring and industrial data management technology, specifically a power transmission disaster prevention monitoring system for industrial data management. Background Technology

[0002] With the increasing demand for disaster prevention and monitoring in ultra-high voltage transmission corridors, industrial data management technology for micro-meteorological complex disaster processes such as icing, strong winds, and galloping has become an important research direction in power transmission safety assurance. In order to achieve continuous perception and risk assessment of the status of key sections of the corridor, it is usually necessary to deploy multiple edge sensing nodes along the target area and build a distributed monitoring link based on edge computing gateways and data processing centers. In the process of power transmission disaster prevention monitoring, existing technologies mainly focus on the collection and uploading of industrial status data itself, such as ice thickness, wind speed and direction, temperature and humidity, vibration and video summaries, while lacking unified quantification and collaborative management of node operating status such as buffer occupation, power supply attenuation, communication retransmission, and task backlog. Especially under the superposition of freezing rain, icing and short-term strong winds, edge nodes are often affected by factors such as limited wind and solar power access, increased link congestion and sudden surge in alarms. Even if some nodes are still online, there may be problems such as increased overall network vulnerability and continuous deterioration of critical data backhaul capability, resulting in limited bandwidth being occupied by redundant data packets, making it difficult to ensure continuous and effective monitoring of critical sections. Therefore, how to effectively characterize the operational vulnerability of edge sensing networks, and judge the speed at which the network approaches failure risk by combining historical trends, and then achieve reasonable switching between regular polling and degraded scheduling, while taking into account the protection of critical nodes, the maintenance of data integrity in non-core areas, and the bandwidth reallocation under alarm storms, has become an urgent technical problem to be solved in the power transmission disaster prevention and monitoring system for industrial data management. Summary of the Invention

[0003] The purpose of this invention is to provide a power transmission disaster prevention and monitoring system for industrial data management, and to solve the following technical problems: To avoid making delayed judgments based solely on coarse-grained information such as whether the network is online, this approach enables proactive identification of instability trends in sensing links. Furthermore, when the network is nearing failure, it ensures the continued availability of communication bandwidth and computing resources for sensing links at critical sections by suspending data sampling tasks in non-core areas.

[0004] The objective of this invention can be achieved through the following technical solutions: A power transmission disaster prevention and monitoring system for industrial data management, comprising: Multiple edge sensing nodes are distributed in the target area. These edge sensing nodes are networked together to form a sensing network. The edge sensing nodes are used to collect industrial status data and node operation status data of the target area. An edge computing gateway communicates with edge sensing nodes and has adjustable transmission bandwidth. The data processing hub communicates with the edge sensing nodes through an edge computing gateway. The data processing center is used to receive industrial status data and node operation status data, and quantitatively calculate the network vulnerability entropy state value of the sensing network based on the node operation status data. The data processing center is also used to calculate the trend of change by combining the stored historical network vulnerability entropy state value sequence, and predict the countdown rate of the sensing network approaching the preset failure time threshold based on the network vulnerability entropy state value and the trend of change, and generate dynamic scheduling instructions based on the countdown rate and the network vulnerability entropy state value. The data processing center is also used to issue dynamic scheduling instructions to edge sensing nodes in order to comprehensively manage the data transmission and computing power status of edge sensing nodes; among them, dynamic scheduling instructions include at least regular polling instructions and suboptimal degradation scheduling instructions.

[0005] Preferably, the node operating status data includes at least cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data; The data processing hub stores a lightweight measurement model; The data processing center is also used to input cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data into the lightweight calculation model, and use the lightweight calculation model to output the network vulnerability entropy state value. The lightweight measurement model is configured to perform feature fusion and uncertainty quantification on cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data to obtain the network vulnerability entropy state value.

[0006] Preferably, the data processing center is also used to compare the network vulnerability entropy state value with a preset fatal threshold: When the network vulnerability entropy state value is greater than or equal to the preset fatal threshold, the data processing center generates a suboptimal degradation scheduling instruction. When the network vulnerability entropy value is less than the preset fatal threshold, the data processing center generates a regular polling instruction; Among them, the regular polling command is used to instruct the edge sensing nodes to perform data collection and transmission according to the preset sleep cycle.

[0007] Preferably, edge-aware nodes are divided into core nodes and non-core nodes; Suboptimal degraded scheduling instructions must include at least data discard instructions and communication cut-off instructions; The data processing center is used to send data discard instructions and communication cut-off instructions to non-core nodes after generating suboptimal degradation scheduling instructions. This instructs non-core nodes to stop sending the industrial status data they collect, i.e. non-core area data, and to enter a deep sleep state where their data acquisition modules are shut down and only their wake-up monitoring circuits are powered. This frees up communication bandwidth for core nodes and preserves the power of core nodes.

[0008] Preferably, the data processing center has a pre-stored generative data inference model, which is a neural network model trained based on historical industrial status data and used to establish the spatiotemporal mapping relationship between core area data and non-core area data. The data processing center is also used to extract the industrial status data, i.e. core area data, that is, after issuing suboptimal and degraded scheduling instructions to non-core nodes. The data processing center is also used to input core area data into the generative data inference model, use the generative data inference model to generate inference fill data based on spatiotemporal mapping relationship, and use the inference fill data to reconstruct data in non-core areas in order to maintain the global integrity of industrial status data.

[0009] Preferably, the industrial status data includes alarm data for sudden anomalies; The data processing center is also used to extract the data features of the sudden abnormal alarm data when the generation rate of the detected sudden abnormal alarm data is greater than the preset alarm storm threshold, and to identify the authenticity of the sudden abnormal alarm data based on the degree of deviation between the data features and the pre-established normal fluctuation range of the equipment status: when the degree of deviation is greater than the preset abnormal threshold, it is determined to be a valid alarm data of a true alarm; otherwise, it is determined to be redundant alarm data of a false alarm. The data processing hub is also used to filter redundant alarm data and reallocate the transmission bandwidth of the edge computing gateway based on the filtered valid alarm data.

[0010] Preferably, the preset failure time threshold is the maximum time span during which the system continuously fails to receive valid data; The data processing center calculates the trend of change and predicts the countdown rate in the following way: it outputs the slope change of the historical network vulnerability entropy state value sequence based on the time series prediction algorithm, and generates the countdown rate by combining it with the preset failure time threshold.

[0011] Preferably, the target area is the ultra-high voltage power transmission corridor area, and the industrial status data is micro-meteorological composite disaster data; The preset failure time threshold is a fixed time threshold pre-stored in the data processing center. This time threshold is calibrated according to the preventive power outage triggering time of the power grid-level security defense system; the edge sensing nodes rely on wind and solar power harvesting devices to obtain electricity.

[0012] Preferably, the data processing center is also used to collect the update node running status data of the edge sensing node after the dynamic scheduling instruction is executed; The data processing center is also used to requantize the network vulnerability entropy state value based on the updated node running status data, and use the requantized network vulnerability entropy state value as a feedback label to iteratively update the internal weight parameters of the lightweight measurement model in order to achieve robust control optimization.

[0013] The beneficial effects of this invention are: 1. This invention obtains the network vulnerability entropy state value by uniformly quantizing the cache, battery and retransmission status, and predicts the countdown rate approaching the failure threshold, thereby generating dynamic scheduling instructions; this mechanism effectively solves the problem of lack of collaborative management of node operation status in existing technologies, and transforms the disaster prevention strategy from passive collection to active risk defense, avoiding the sudden interruption of the sensing link in a disaster. 2. This invention divides the network into core and non-core nodes, and triggers suboptimal degradation scheduling when the network vulnerability reaches a critical threshold, actively cutting off the communication of non-core nodes and putting them into deep sleep. This mechanism effectively solves the problem of limited resources being preempted by redundant data packets, and concentrates communication bandwidth and power resources on key sections, ensuring the reliable transmission of core disaster prevention data. 3. This invention introduces a generative data extrapolation model. During the degrade and silent period of non-core nodes, extrapolation and filling data are generated by combining the data transmitted back in real time by core nodes with historical spatiotemporal mapping relationships. This mechanism effectively makes up for the problem of blind spots in regional situation caused by cutting off non-core nodes. While shrinking communication resources, it reconstructs non-core area data and maintains the integrity of global disaster monitoring. 4. When the generation rate of sudden abnormal alarms exceeds the storm threshold, this invention identifies the authenticity of alarms based on the degree of deviation between data characteristics and the normal fluctuation range, and directly filters redundant alarms. This mechanism effectively solves the problem of network congestion caused by a large number of invalid alarms under micro-meteorological composite disasters, and quickly reallocates the edge gateway bandwidth to core valid data to avoid delays in critical decisions. 5. This invention collects the updated node running status data after executing dynamic scheduling instructions, and uses this data to requantize the status value as a feedback label to iteratively update the internal weight parameters of the lightweight calculation model. This mechanism enables network vulnerability calculation to no longer be limited by static initial settings, and can continuously adaptively correct itself based on real feedback, thus achieving model optimization that is more in line with the changing conditions on site. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1This is a schematic diagram of a power transmission disaster prevention monitoring system for industrial data management provided in an embodiment of this application. Detailed Implementation

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

[0016] Please see Figure 1 A power transmission disaster prevention and monitoring system for industrial data management, comprising: multiple edge sensing nodes distributed in the target area, the multiple edge sensing nodes being networked to form a sensing network, the edge sensing nodes being used to collect industrial status data and node operation status data of the target area; An edge computing gateway communicates with edge sensing nodes and has adjustable transmission bandwidth. The data processing hub communicates with the edge sensing nodes through the edge computing gateway. The data processing hub is used to receive industrial status data and node operation status data, and quantitatively calculate the network vulnerability entropy state value of the sensing network based on the node operation status data. The data processing center is also used to calculate the trend of change by combining the stored historical network vulnerability entropy state value sequence, and predict the countdown rate of the sensing network approaching the preset failure time threshold based on the network vulnerability entropy state value and the trend of change, and generate dynamic scheduling instructions based on the countdown rate and the network vulnerability entropy state value. The data processing center is also used to issue dynamic scheduling instructions to edge sensing nodes in order to comprehensively manage the data transmission and computing power status of edge sensing nodes; among them, dynamic scheduling instructions include at least regular polling instructions and suboptimal degradation scheduling instructions.

[0017] This embodiment provides a power transmission disaster prevention monitoring mechanism for industrial data management. Specifically, this embodiment takes the continuous disaster process of UHV transmission corridors entering freezing rain and icing accompanied by short-term strong winds as the unified main scenario. Multiple edge sensing nodes are set up along the line. Each edge sensing node is installed on towers, high-risk areas of conductor icing, valley wind gaps, or micro-topographic change areas to collect industrial status data such as conductor galloping, icing thickness, wind speed and direction, temperature and humidity, tower base vibration, and video clip summaries. At the same time, it collects node operation status data such as buffer usage, remaining power supply capacity, communication quality, and task queue length. Multiple edge sensing nodes are connected to the edge computing gateway through a self-organizing network, and the edge computing gateway is then connected to the data processing center to form an integrated data management link from end to edge to cloud. In detail, the data processing center does not regard each node as merely a simple source of data, but further views the entire network as an industrial sensing system that becomes increasingly vulnerable as the disaster evolves; The so-called network vulnerability entropy state value corresponds to a comprehensive characterization of the degree of instability accumulation in engineering. It does not reflect the failure of a single node, but rather the risk level of the entire sensing network evolving into a state of data link blockage under the combined effects of energy, buffering, communication retransmission, and task backlog. For example, if the ice-covered node on the conductor is still measurable, but its buffer is close to saturation, the link retransmission is frequent, and the power supply capacity of adjacent relay nodes is reduced, then although the node has not lost contact, the sustainable backhaul capability of the entire critical section is still in a deteriorating state. The data processing center quantifies this deteriorating state based on real-time received node operation status data; at the same time, it identifies whether the state value is a temporary fluctuation or a continuous deterioration by combining the historical change trajectory of this state value. The data processing center further calculates how quickly the current network will approach the failure time threshold that triggers power grid protection actions due to continuous failure to receive valid data, and generates dynamic scheduling instructions accordingly. In a simplified data flow analysis model, four nodes near a critical section can be abstracted as nodes. ,node ,node and nodes ,in and Located in a high-risk area for icing, Located in the Yamaguchi wind tunnel area, Located in a low-risk blind spot; At some point, and Continuous reporting of icing and dancing data, Report wind disturbance data. Report routine environmental data; The data processing center will combine the node operation status data of these four nodes into a system containing... state, state, status and Current periodic state fragment Then, compared with the state fragment of the previous cycle. State segments of the first two cycles Composition of time series; If the sequence shows a continuous increase in buffer capacity, a continuous increase in retransmissions, and a continuous decrease in power supply capacity, it indicates that the network is not experiencing instantaneous congestion, but is approaching an unsustainable state. In this case, the data processing center generates dynamic scheduling instructions that are more inclined to preserve the survivability of the links. If the sequence only shows short-term spikes and falls, then regular polling scheduling can still be maintained. Regarding the anomaly handling mechanism, if a node only has industrial status data and temporarily lacks node operation status data, the data processing center can use the node's most recent valid operation status data as a short-term placeholder reference, but at the same time reduce the weight of the node in the overall network status quantification to avoid misleading scheduling by old status. If the backhaul link between the edge computing gateway and the data processing center is interrupted for a short time, the edge computing gateway can maintain local execution for a predetermined protection window according to the most recently received scheduling rules. If uplink communication cannot be restored within the protection window, the edge computing gateway will switch to conservative mode and prioritize the retention of simplified monitoring data for critical sections. If individual nodes report obviously abnormal zero values, full values, or data with incorrect format, a validity check will be performed first. The data will not be included in the current period's vulnerable state quantification, but the original message will be retained for subsequent fault analysis. In a cross-mountain ultra-high-voltage power transmission corridor, the system normally operates using an energy-saving strategy of long hibernation and short wake-up; however, when freezing rain occurs at night and is compounded by crosswinds in the canyon, power outages occur at critical sections. , At the same time, increased icing was detected on the conductor. Increased gusts were detected; At this time, the normal data collection rhythm was disrupted by a large number of sudden alarms. If all nodes are still woken up and sent back with the same frequency and priority according to the static rules, non-core background monitoring data will compete with key icing data for limited bandwidth. The data processing center identifies, through the network vulnerability entropy status value, that although there are still nodes online, the link has shown signs of overall instability. It then generates dynamic scheduling instructions and sends them to the edge computing gateway and related nodes to comprehensively manage data transmission and computing power status. The purpose of this step is to upgrade the single disaster prevention monitoring method that focuses only on static results to a dynamic industrial data management method that simultaneously assesses the remaining stable operating cycle of the system. This will enable the early identification of the instability trend of the sensing link and provide a unified decision-making basis for subsequent routine polling and suboptimal degradation scheduling.

[0018] In a preferred embodiment of the present invention, the node operating status data includes at least cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data; the data processing center stores a lightweight calculation model. The data processing center is also used to input cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data into the lightweight calculation model, and use the lightweight calculation model to output the network vulnerability entropy state value. The lightweight measurement model is configured to perform feature fusion and uncertainty quantification on cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data to obtain the network vulnerability entropy state value.

[0019] This embodiment provides a lightweight network vulnerability entropy measurement mechanism; specifically, in the previous embodiment, if the network status is judged solely based on whether a node is online or whether there is data feedback, it is easy to overlook the hidden failure process where the node's communication connection is normal but its internal operating indicators show a deteriorating trend. For example, nodes may still be reporting data, but internal caches may be piling up, power supply components may have reduced output capacity after multiple deep charging and shallow discharging cycles, and links may be frequently retransmitted after a disaster. These phenomena will amplify the risk of instability in the following cycles. Therefore, this embodiment further introduces cache queue occupancy rate data, battery health decay data, and communication data packet retransmission rate data, and outputs network vulnerability entropy status value through a lightweight calculation model; Specifically, cache queue occupancy reflects the extent to which data that cannot be sent out in time is accumulating within a node; In the context of power transmission disaster prevention, cache backlog means that newly generated high-time-sensitive data such as icing and galloping may be squeezed out by old data, thus delaying the delivery of critical risks. Battery health degradation reflects the sustainability of the node power supply system. Especially when wind and solar power extraction is blocked by freezing rain, even if the node still has voltage output for a short time, it may be unable to withstand the pulse load caused by frequent wake-up and transmission due to cell aging. The retransmission rate of communication data packets directly reflects the difficulty of effective transmission of wireless links under the conditions of wind and rain attenuation, obstruction, multipath interference or relay congestion. The lightweight measurement model maps the three types of state variables to a comparable range and then uses their distribution dispersion to measure the cumulative level of uncertainty within the network. The feature fusion and uncertainty quantification process employs the negative logarithmic product algorithm to characterize the positive correlation between the dispersion of state variables and the probability of unsustainable network operation. Simultaneously, to ensure mathematical and engineering rigor and avoid situations where a state, such as an extremely low retransmission rate, has a normalized probability distribution value of zero, which could lead to meaningless or infinitely large errors during negative logarithmic calculations and system crashes, the model introduces a very small positive bias as a smoothing constant for fault-tolerant computational protection. The specific formulas for feature fusion and uncertainty quantification are as follows: in, To quantify the calculated network vulnerability entropy state values, The internal weight parameters for the operational status data of each node. The probability distribution values ​​corresponding to the input state data. The bias is a very small positive number; In addition, in order to meet the needs of subsequent system optimization and control, the lightweight calculation model has preset internal weight parameters for the running status data of each node. Before obtaining the probability distribution value, the initial state quantity is weighted and mapped using the internal weight parameters, which provides a basis for adjusting the influence ratio of each state indicator in the future. To facilitate understanding, a simplified analysis model can be constructed; assuming that only three types of state segments are observed within a certain period: the first segment representing the cache state The second segment characterizing the health status of the power supply And the third segment representing the retransmission state ; like , , Of the three nodes, most of the nodes Both are close to low usage. All are in good health If all retransmissions are at a low retransmission rate, the state distribution obtained by the model is more concentrated, indicating that the overall network operation order is relatively stable. Conversely, if High cache Significant power attenuation occurred. When high retransmission occurs, the three types of states spread in different directions of instability, and the vulnerability entropy state value output by the model is higher, indicating that although the network has not been completely interrupted, it has already shown a multi-source instability trend. The reason why this model adopts a lightweight implementation is that during a disaster, the scheduling logic must be faster than the link deterioration speed, and it cannot rely on a complex and time-consuming model to give results. Regarding the exception handling mechanism, if only cached and retransmitted data are obtained in a certain period, and the battery health degradation data fails to be updated due to sampling module maintenance or short-term anomaly, the battery health baseline stored in the previous maintenance period can be called to participate in the current calculation, and the result is marked as a partial observation state for subsequent scheduling to use with caution. If any of the three states shows a significant over-range value or a physically impossible jump, such as an unreasonable and large rebound in battery health within a very short period of time, it is first determined to be a sampling anomaly, and the item is replaced with a smoothed nearest neighbor reference value. If multiple consecutive cycles are abnormal, a node status abnormality flag is triggered, and the data processing center reduces its dependence on that node in the scheduling. If all three states are missing, the vulnerability assessment result of the node will not be updated in this period, and the node will be directly included in the potentially high-risk set. In the aforementioned incident of freezing rain at night along the mountain power transmission corridor Although it still transmits icing data periodically, its cache queue has gradually increased from low usage to near saturation. Due to continuous rain and ice cover, the effective discharge capacity of the energy storage unit has decreased significantly. Due to the deterioration of the electromagnetic environment in the Yamaguchi wind tunnel area, the retransmission rate has increased rapidly. After the data processing center inputs these three states into the lightweight calculation model, it finds that the operation of the entire network has evolved from uniform tension to multi-directional instability, thus obtaining a high network vulnerability entropy state value; at this point, even if no node has completely gone offline, there is already a basis for adjusting the scheduling in advance. The purpose of this step is to unify the three different types of hidden risks—cache backlog, power supply aging, and communication conflicts—into a lightweight metric that can be updated in real time. This enables early identification of potential instability in the sensing network and avoids making delayed judgments based solely on coarse-grained information such as whether the network is online.

[0020] In a preferred embodiment of the present invention, the data processing center is further configured to compare the network vulnerability entropy state value with a preset fatal threshold: when the network vulnerability entropy state value is greater than or equal to the preset fatal threshold, the data processing center generates a suboptimal degradation scheduling instruction. When the network vulnerability entropy value is less than the preset fatal threshold, the data processing center generates a regular polling instruction; the regular polling instruction is used to instruct the edge sensing nodes to perform data collection and transmission according to the preset sleep cycle.

[0021] This embodiment provides a mechanism for switching scheduling modes based on fatal thresholds. Specifically, in the previous embodiment, although the network vulnerability level can be quantified, if the system only outputs a risk value and lacks a clear scheduling switching threshold, it is still difficult to form a unified and timely control action in actual engineering. Especially in power transmission disaster prevention scenarios, edge nodes mostly rely on wind and solar energy harvesting devices that are significantly affected by weather. If the scheduling switch is delayed, it is easy to exceed the system's maximum tolerable latency between continuing to pursue full data and timely protection of critical links. Therefore, this embodiment achieves a clear switch between regular polling instructions and suboptimal degraded scheduling instructions by comparing the network vulnerability entropy state value with a preset fatal threshold. Specifically, the regular polling command is applicable when the network is still in the sustainable working range. At this time, the system controls each node to wake up, collect data, upload data, and then enter sleep mode according to a preset sleep cycle, so as to reduce unnecessary continuous power supply and radio frequency occupation. In ultra-high voltage transmission corridors, this strategy is usually used in stable weather or periods of mild disturbance to maintain basic coverage with low energy consumption; however, when the network vulnerability entropy value is greater than or equal to the preset fatal threshold, it indicates that the network is no longer suitable for adhering to the original balanced acquisition strategy. At this point, the generation of a suboptimal degraded scheduling instruction means that the system determines that continuing to maintain the same level of fine monitoring across the entire network would lead to network overload, thus triggering a degraded scheduling mechanism that prioritizes maintaining the continuity of core data at critical sections. The switching logic can be illustrated by a simplified analysis model; assume that a certain power transmission corridor has a first node group A, a second node group B, and a third node group C; normally A is responsible for icing, B is responsible for wind deflection, and C is responsible for environmental coverage. If the system detects that the network vulnerability entropy value is consistently below the preset fatal threshold for multiple consecutive cycles, it will uniformly issue regular polling instructions to A, B, and C, such as maintaining the established sleep-wake rhythm. When the disaster escalates, the buffering, power supply and retransmission status of the nodes around A and B deteriorate at the same time. When the network vulnerability entropy value touches or exceeds the preset fatal threshold, the system no longer maintains the original rhythm, but switches to degrade mode. The term "suboptimal" here does not mean "inferior performance," but rather that, relative to the optimal collection of all data, the priority has shifted to overall survival and the continuity of key data. Regarding the anomaly handling mechanism, if the network vulnerability entropy state value happens to be near the preset fatal threshold and fluctuates repeatedly in adjacent periods, a hysteresis interval or a minimum hold window can be set to avoid frequent jittering of scheduling instructions. In other words, once the suboptimal degradation mode is entered, it should be maintained for at least one protection period until the status is continuously restored before exiting. If the current network vulnerability entropy state value is not reliable due to the lack of data from some nodes, the data processing center can adopt a conservative control principle that prioritizes system stability and use a more conservative scheduling strategy for key sections. If the system detects that the preset fatal threshold is no longer suitable for the current season, line structure, or equipment aging level, it can be recalibrated offline by the operation and maintenance backend and then updated, avoiding frequent modifications during the online phase of a disaster. In the aforementioned scenario of freezing rain combined with strong winds, the system normally adopts a regular polling rhythm of waking up once every 12 minutes. In the early stage of the disaster, although the number of alarms increases, the network vulnerability entropy value has not yet reached the preset fatal threshold, and the data processing center still maintains this rhythm to balance coverage and energy consumption. As icing intensifies, link retransmissions increase, and energy storage capacity decreases, the network vulnerability entropy value becomes greater than or equal to the preset fatal threshold. The data processing center immediately stops pursuing full network synchronization and complete backhaul, and instead generates suboptimal degraded scheduling instructions to create conditions for subsequent bandwidth release and power preservation. The purpose of this step is to provide an actionable control exit for the risk quantification results, thereby achieving a closed-loop connection from risk identification to triggering actions and preventing the network from continuing to use unsuitable conventional data collection strategies during high-risk phases.

[0022] In a preferred embodiment of the present invention, the edge sensing nodes are divided into core nodes and non-core nodes; the suboptimal degradation scheduling instruction includes at least a data discard instruction and a communication cut-off instruction; the data processing center is used to send the data discard instruction and the communication cut-off instruction to the non-core nodes after generating the suboptimal degradation scheduling instruction, so as to instruct the non-core nodes to stop sending the industrial status data they collect, i.e., non-core area data, and enter a deep sleep state in which their data acquisition modules are turned off and only their wake-up monitoring circuit is powered, thereby releasing the communication bandwidth for the core nodes and preserving the power of the core nodes.

[0023] This embodiment provides a core node preservation suboptimal degradation scheduling mechanism. Specifically, in the previous embodiment, although it was possible to switch between the normal polling and degradation modes, if an indiscriminate scheduling mechanism is still used for all nodes after entering the degradation mode, and only slight rate limiting or uniform frequency reduction is performed, then when the disaster continues to deepen, the entire network may still fall into a data failure state due to communication overload caused by all nodes equally competing for resources. Therefore, this embodiment further divides the edge sensing nodes into core nodes and non-core nodes, and performs a stronger contraction action on the non-core nodes in the degradation mode; Specifically, core nodes are usually deployed in key transmission sections, areas prone to icing, areas sensitive to tower stress, areas with overlapping wind zones, or areas most closely linked to power grid protection. If the data from these nodes is interrupted, the scheduling end will directly lose the basis for judging critical risks; non-core nodes mostly undertake background blind spot filling, spatial interpolation assistance, or routine perception tasks in low-risk areas. The data collection priority of these nodes is low, and their data uploading process can be actively stopped when the available communication bandwidth is less than the set threshold. Once the data processing center generates a suboptimal degradation scheduling instruction, it issues data discard instructions and communication cut-off instructions to non-core nodes, causing them to stop sending data to non-core areas, shut down high-energy-consuming data acquisition modules, and retain power to the wake-up monitoring circuit so that they can reconnect to the network when the risk decreases. In essence, this is to transfer limited bandwidth, transmission opportunities and remaining power from supplementing information at the edge to continuous information at the critical section. In the simplified analysis model, nodes can be divided into those containing nodes. and core set and containing nodes , and Non-core sets ; When the system enters suboptimal degradation mode , , It no longer uploads the collected regular environmental segments and low-priority alarm segments, and its sampling, encoding and transmission tasks are all stopped. Only a low-power listening unit is retained to receive subsequent recovery commands. and Then continue to transmit key data related to icing, conductor galloping, and tower stress according to the compressed high-priority scheme; In this way, the data flow that was originally queued together at the gateway is significantly reduced, and the packets of key nodes have a higher probability of being sent and a shorter waiting time; in terms of the exception handling mechanism, the division between core nodes and non-core nodes is not permanent. If a non-core area evolves into a new risk center during a disaster, such as when an abnormal sway of a conductor suddenly appears in a valley crosswind area, the data processing center can dynamically increase its priority and move it from the non-core set to the core set. If a core node experiences a power supply failure or hardware malfunction during execution, a nearby node can be temporarily selected as a backup core node. For non-core nodes that have been disconnected from communication, the wake-up listening circuit is still retained to prevent them from losing network access opportunities; once the risk is mitigated or the operation and maintenance department issues a reconnection command, such nodes can be restored in batches to prevent instantaneous reconnection from causing congestion again. If the information reported by a non-core node is not continuous monitoring master data, but contains security messages such as local self-test failure, battery swelling symptoms or tower intrusion, then such messages are not subject to the ordinary data discarding rules and can be sent as exception messages. In the aforementioned scenario of freezing rain and strong winds, crossing the main span of the mountain pass... and Designated as a core node, located on the outer gentle slope and in the background monitoring zone. , , Set as a non-core node; When an alarm storm leaves only a small amount of bandwidth remaining on the gateway, the data processing center directs... , , When a shrink command is issued, they stop routine environmental data collection and transmission, enter deep sleep, and retain only the listening circuits that receive wake-up commands; thus, and Key data on ice thickening and conductor galloping are continuously delivered, preventing continuous data gaps at critical sections during the most dangerous moments. The purpose of this mechanism is to sacrifice local information coverage in exchange for the continued survival of critical segment sensing links when the network is about to fail, thereby achieving a concentrated allocation of bandwidth, power, and transmission time slots to core nodes.

[0024] In a preferred embodiment of the present invention, the data processing center pre-stores a generative data inference model, which is a neural network model trained based on historical industrial status data and used to establish a spatiotemporal mapping relationship between core area data and non-core area data. The data processing center is also used to extract the industrial status data, i.e. core area data, that is, after issuing suboptimal and degraded scheduling instructions to non-core nodes. The data processing center is also used to input core area data into the generative data inference model, use the generative data inference model to generate inference fill data based on spatiotemporal mapping relationship, and use the inference fill data to reconstruct data in non-core areas in order to maintain the global integrity of industrial status data.

[0025] This embodiment provides a data reconstruction mechanism for non-core areas after degradation; specifically, in the previous embodiment, cutting off communication of non-core nodes can significantly preserve critical links, but this also brings new problems: although the dispatch center preserves the continuity of data at critical sections, it loses the global observation of the disaster expansion trend in the surrounding areas. If non-core area information is completely abandoned, subsequent disaster propagation analysis, regional linkage judgment, and post-mortem assessment will be affected. Therefore, this embodiment further introduces a generative data extrapolation model, which uses core area data to extrapolate, fill in, and reconstruct non-core area data during the non-core node silence period. In detail, the micro-meteorological and structural responses of ultra-high voltage transmission corridors are not isolated from each other; in historical operation, it can usually be observed that after a strong wind intensifies in a certain mountain pass area, the conductor galloping at the downwind slope node will respond after a certain period of time. There is a spatial transmission relationship between the temperature and humidity changes in a certain high-altitude icing zone and the icing growth in the adjacent gentle slope area; the stress changes of adjacent tower sections on the same line during severe weather also have a relatively stable spatiotemporal correlation; the generative data extrapolation model is established based on this historical spatiotemporal mapping relationship; When non-core nodes are downgraded and silenced, the data processing center extracts the core area data transmitted back in real time from the core nodes, inputs it into the pre-stored model, and generates inference and filling data on the state of non-core areas, thereby maintaining the continuity of the global data view. In a minimal analysis model, fragments can be extracted. and core area fragment and containing fragments , and Non-core region fragments Historical data shows that when The icing continued to thicken and When crosswinds intensify, Humidity often rises in the later stages of the day. Slight dancing movements appeared. Maintain relative stability; During the disaster, , , It has been silenced; the data processing center has only received [data]. and Real-time data; generative data inference models output inference fragments based on this data. , , As the result of the simulation, it is labeled as reconstructed data rather than the original measured data; in this way, the dispatcher can still see the regional situation distribution in the map interface or analysis module, thus avoiding data gaps in the global state mapping. Regarding the anomaly handling mechanism, the inference and filling data must be subject to credibility constraints; if the core area data itself has obvious missing, conflict, or quality degradation, the inference for areas with weak correlation will be suspended, and only the local reconstruction with high confidence will be retained. If the historical correlation between non-core areas and core areas no longer holds under the current season, wind direction, or terrain humidity conditions, the model output should be downgraded and, if necessary, displayed directly as an unknown area instead of being forcibly filled in. If some of the silent nodes periodically recover a heartbeat message, then this type of sparse real data can be used to correct the generated results and avoid long-term drift. For hard constraint parameters that are directly used in power grid strong protection decisions, real core data should be used first, and reconstructed data should only be used as auxiliary background and should not replace key protection criteria. During the operation of the aforementioned main cross-section link, the outer gentle slope area , , It has entered deep hibernation; the data processing center continues to receive data. and The data on icing, galloping, and local wind fields were fed into a generative data extrapolation model that had been trained based on years of corridor disaster samples. The model is reconstructed accordingly. , , The non-core area data corresponding to the region allows dispatchers to see whether the peripheral area is still in a state of damp cold accumulation, mild tremors, or relatively stable; as a small number of recovered heartbeat messages arrive, the central system then corrects these reconstruction results; The purpose of this mechanism is to maintain the continuous expression of the overall disaster situation as much as possible after actively cutting off non-core acquisition links, thereby achieving a balance between preserving key measured data and avoiding the collapse of the overall view.

[0026] In a preferred embodiment of the present invention, the industrial status data includes sudden abnormal alarm data; the data processing center is also used to extract data features of the sudden abnormal alarm data when the generation rate of the detected sudden abnormal alarm data is greater than a preset alarm storm threshold, and to identify the authenticity of the sudden abnormal alarm data based on the degree of deviation between the data features and the pre-established normal fluctuation range of the equipment status. When the deviation is greater than the preset abnormal threshold, it is determined to be a valid alarm data for a true alarm; otherwise, it is determined to be redundant alarm data for a false alarm. The data processing hub is also used to filter redundant alarm data and reallocate the transmission bandwidth of the edge computing gateway based on the filtered valid alarm data.

[0027] This embodiment provides an alarm storm filtering and bandwidth reallocation mechanism; specifically, in the aforementioned main scenario, when freezing rain and icing are superimposed with short-term strong winds, a large number of sensors are often triggered simultaneously in a short period of time. If all alarms are uploaded with high priority, the gateway's limited bandwidth will be quickly exhausted by a massive number of duplicate, concurrent, or falsely triggered alarms, which will prevent the key data that truly determines the protection action from passing through in a timely manner. Therefore, this embodiment identifies sudden abnormal alarm data in industrial status data, and when its generation rate exceeds the preset alarm storm threshold, it performs authenticity verification and bandwidth reallocation for alarms. Specifically, a true alarm refers to an alarm whose data characteristics are consistent with the actual abnormal mechanism of the line and have a sufficiently significant deviation from the normal fluctuation range; for example, increased conductor icing is often accompanied by synergistic features such as changes in temperature and humidity, changes in clamp stress, and increased low-frequency swing amplitude. False alarms caused by sudden electromagnetic interference, local equipment burrs, or wind noise usually only show short-term spikes in a single channel and lack multi-source consistency. After the data processing center detects that the alarm generation rate is higher than the alarm storm threshold, it extracts the data characteristics of each alarm, including duration, amplitude change pattern, whether spatially adjacent nodes respond in coordination, and whether it is consistent with the meteorological background, and then judges the deviation from the preset normal fluctuation range. If the deviation is significant enough and conforms to the physical correlation law, it is determined to be valid alarm data; otherwise, it is determined to be redundant alarm data. The system filters redundant alarms and prioritizes the transmission bandwidth of the edge computing gateway to be allocated to valid alarms and key monitoring data. This can be illustrated using a microscopic analysis model; assuming that within a certain time period, , , A total of six alarms were generated, which are recorded as alarms. to ;in , From and This manifests as adjacent nodes simultaneously experiencing increased icing and enhanced dancing within the same wind gap area; From This manifests as a single-point instantaneous vibration peak; to This refers to the same minor deviation repeatedly reported by multiple nodes within a very short period of time. After the data processing center extracts the persistence, proximity consistency, and background matching of these alarms, it can... , The alarm is deemed valid and will be... to The alarm was flagged as redundant; therefore, the gateway stopped using it. to Continuously occupying upload resources, while allocating more time slots to , And the related original supporting data; Regarding the anomaly handling mechanism, if the alarm generation rate does not exceed the alarm storm threshold, the system may not activate strong filtering and will only process according to the regular priority rules to avoid excessive suppression of information during stable periods. If an alarm deviates from the threshold and is difficult to determine directly, it can be temporarily classified as an alarm pending confirmation, and only the summary is uploaded. It can be upgraded or downgraded after receiving supplementary information from adjacent nodes. If it is difficult to complete the identification due to the lack of some background data, alarms involving key sections should be retained and compressed rather than discarded directly. If the same type of alarm occurs simultaneously on multiple non-core nodes but the core nodes do not respond, then when the non-core nodes have been downgraded, only a small number of sample packets with the strongest spatial representativeness can be retained to avoid sending all packets and causing secondary congestion. During the aforementioned nighttime disaster in the mountain corridor, the system received a large number of alarms related to abnormal vibration, abnormal displacement, and abnormal temperature within the preset time window. A significant portion of these alarms originated from wind noise, short-term collisions, or repeated triggering. The data processing center discovered that the alarm generation rate was significantly higher than usual and thus initiated the alarm storm handling process. After feature extraction, only alarms that match the thickening of ice and the continuous crosswinds and show proximity consistency in the key sections are retained as valid alarms; other short-term anomalies lacking coherent evidence are filtered out; the edge computing gateway then reallocates its limited bandwidth to the ice data, dancing data and valid alarm summaries of the key sections. The purpose of this mechanism is to suppress the excessive consumption of link communication resources by redundant messages when alarms occur in a concentrated manner, thereby realizing the targeted allocation of bandwidth resources to effective risk data and avoiding redundant alarm data from interfering with the normal defense decision-making of the system.

[0028] In a preferred embodiment of the present invention, the preset failure time threshold is the maximum time span during which the system continuously fails to receive valid data; the specific method for the data processing center to calculate the change trend and predict the countdown rate is as follows: based on the time series prediction algorithm, the slope change state of the historical network vulnerability entropy state value sequence is output, and the countdown rate is generated in combination with the preset failure time threshold.

[0029] This embodiment provides a failure threshold approach rate prediction mechanism. Specifically, in the aforementioned implementation, the system is already able to identify the current network vulnerability and switch the scheduling mode accordingly. However, if the evaluation is based solely on the network vulnerability at the current discrete moment, it is easy to overlook the trend of the system accelerating its deterioration toward the failure boundary. The core technical challenge of power transmission disaster prevention systems during disasters lies not in handling sudden surges in data, but in preventing communication data gaps at critical sections that exceed the tolerance threshold. Once such a continuous gap exceeds the tolerance range of the power grid defense system, it may trigger preventive protection actions. Therefore, this embodiment further introduces a preset failure time threshold and its countdown rate prediction. Specifically, the preset failure time threshold represents the maximum time span that the system can tolerate when it fails to receive valid data for an extended period of time. Its source can be determined by the triggering rules on the power grid security defense side. The data processing center does not simply record how long it has been since it received data, but rather combines historical network vulnerability entropy state value sequences to observe how quickly the network approaches a preset fatal threshold; if the network vulnerability entropy state value is high but the growth slope is slowing down, it indicates that the current scheduling measures may be working. If the network vulnerability entropy value continues to rise at an accelerating rate, it indicates that instability has not yet been suppressed and the system is not far from a true data blind zone. The so-called countdown rate can be understood in engineering as the urgency with which the system approaches the preset failure time threshold. The higher the rate, the more necessary it is to take decisive conservative scheduling actions. In the simplified analysis model, network vulnerability entropy state value segments can be obtained for four consecutive periods, denoted as segments. fragments fragments and fragments ;like arrive A slow increase indicates that while the network is deteriorating, there is still room for adjustment; if... arrive The sharp increase indicates that the impact of the disaster and the deterioration of the supply chain are amplifying each other; The data processing center uses a time series prediction algorithm to calculate the growth slope of the historical network vulnerability entropy state value sequence as the trend. Then, it calculates how far the current network vulnerability entropy state value is from the preset fatal threshold, i.e., the difference, and combines it with the growth slope to deduce the remaining time, i.e., the ratio of the difference to the growth slope. The ratio of the preset failure time threshold to the remaining time is calculated into a countdown rate that approximates the preset failure time threshold; This emphasizes trend assessment rather than static point judgment, allowing scheduling instructions to react more proactively. In this calculation principle, to prevent serious division errors caused by a zero or negative growth slope, the data processing center incorporates explicit fault-tolerant calculation protection logic. If the calculated growth slope is determined to be less than or equal to the set minimum threshold for safety approaching zero, which indicates that network deterioration has stopped or is in an improving state, the system will directly terminate the process of using the slope as a division to derive the remaining time, and instead set the remaining time to the maximum limit value representing safety, and make the calculated countdown rate zero, thereby avoiding system calculation abnormalities caused by the denominator being zero. Regarding anomaly handling mechanisms, if the historical network vulnerability entropy state value sequence is not long enough, for example, when the system has just been deployed or a large amount of data is missing for a certain period of time, a trend estimation with a shorter time window can be used, and the result can be marked as low confidence. If a single abnormal spike appears in the network vulnerability entropy state value, but recovers rapidly in adjacent periods, smoothing or outlier removal should be used to avoid misjudging sporadic noise as continuous deterioration. If the trend of the growth slope is close to zero or turns negative, i.e., declining, it indicates that the previous scheduling may have released the pressure. At this time, the system can maintain the existing strategy and avoid excessive contraction. If the estimated remaining time is shorter than the minimum control cycle required to complete a full dispatch and execution, the central control should directly select the most conservative emergency mode without waiting for further observation. In the aforementioned main scenario, the system is already close to the high-risk edge of the 12-minute sleep rhythm, while the maximum time tolerance for continuous valid data loss allowed by the power grid protection side is 15 minutes. When freezing rain and strong winds combine, the network vulnerability entropy state value over several historical cycles not only continues to rise, but also the growth slope becomes increasingly steep. Based on this calculation, the data processing center determined that if the original strategy continued to be followed, the critical section was very likely to approach the preset failure time threshold in a very short time. Therefore, even if individual node messages are still occasionally received, it is already regarded as a state of rapid approach to the threshold, and stronger scheduling protection is triggered in advance, rather than passively dealing with it after a real interruption. The purpose of this mechanism is to shift the judgment of whether failure will occur to the judgment of how quickly failure is approaching, thereby achieving time-based management of data blind spot risks and enhancing the lead time of scheduling actions.

[0030] In a preferred embodiment of the present invention, the target area is an ultra-high voltage transmission corridor area, and the industrial status data is micro-meteorological composite disaster data; the preset failure time threshold is a fixed time threshold pre-stored in the data processing center, and the time threshold is calibrated according to the preventive tripping and network disconnection triggering time of the power grid-level security defense system; the edge sensing node relies on wind and solar energy harvesting devices to obtain electricity.

[0031] This embodiment provides a scenario-based deployment mechanism for micro-meteorological complex disasters in ultra-high voltage transmission corridors; specifically, the aforementioned implementation methods have provided general methods for network vulnerability identification, threshold switching, node degradation, data reconstruction, and alarm storm handling; This embodiment further limits its applicable scenario to the UHV transmission corridor area, the industrial status data to micro-meteorological composite disaster data, and the edge sensing nodes to be powered by wind and solar energy harvesting devices, so that the entire technical solution closely corresponds to the actual power grid disaster prevention conditions. In detail, ultra-high voltage power transmission corridors are characterized by long distances across regions, large terrain spans, strong meteorological differences, and difficulties in manual inspection; Especially in high-altitude areas, mountain passes, canyons, ice-covered zones, and dancing zones, disasters are not caused by a single factor, but often manifest as a complex process involving multiple factors such as temperature, humidity, precipitation phase, crosswind, conductor attachment, and mechanical vibration. Therefore, industrial condition data preferably includes local wind speed and direction, temperature and humidity, rain and snow phases, icing changes, tower tilt, conductor galloping, sag changes, insulator string status, and video or image summaries, etc. On the other hand, edge sensing nodes rely on wind and solar power devices to obtain electricity. When the solar radiation and wind parameters meet the preset threshold, they maintain a positive power balance. However, during periods of continuous overcast and snowy weather, icing and shading, or low wind speed, they will enter a long-term net power consumption state. Therefore, the aforementioned dispatch mechanisms are not optional, but necessary conditions for maintaining the continuity of disaster prevention and monitoring under the premise of naturally limited energy resources. In the simplified analysis model, the corridor can be divided into high-risk zones. transition zone and background band ; Deploy high-density core nodes to focus on monitoring icing, galloping, and stress. Deploy auxiliary nodes to focus on monitoring wind field transmission and the evolution of wet and cold conditions; Deploy low-density blind spot detection nodes; The data processing center pre-stores a preset failure time threshold, i.e., a fixed time threshold. This time threshold is calibrated based on the maximum allowable gap between the loss of effective monitoring data and the triggering of preventive power outages by the power grid-level security defense system. When the aforementioned vulnerability identification and scheduling logic is running, the preset failure time threshold is invoked as a unified time boundary, without the need for temporary manual setting at the disaster site; Regarding the anomaly handling mechanism, although the fixed time threshold corresponding to the preset failure time threshold is stored in advance, it can be updated offline in the operation and maintenance window according to the protection strategy adjustment. It is not recommended to frequently change it during a disaster to avoid inconsistent decision-making standards within the same event. If some edge nodes are equipped with backup batteries or short-term external power supply, they can still be included in the same scheduling framework, but their power supply risk weight can be appropriately reduced when vulnerability assessment. If local areas cannot stably utilize wind and solar energy due to terrain or maintenance conditions, a hybrid power supply node and wind and solar nodes can be networked together. The system will still be uniformly scheduled according to the node's operating status without changing the overall approach. In an ultra-high-voltage transmission corridor traversing high-altitude mountainous areas and windy valleys, The main area was severely affected by ice accumulation. This is the transition zone between adjacent slopes and air ducts. For the outer background area; When winter disasters occur, and The solar panels on the nodes were blocked by ice, and the wind turbines were affected by unstable gusts, resulting in a significant decrease in the energy replenishment capacity. Because the data processing center pre-stores a preset failure time threshold determined by the preventive tripping and network disconnection trigger time calibrated by the power grid defense system, the system can always perform vulnerability identification and countdown rate calculation and scheduling decisions around this unified preset failure time threshold during a disaster, without the need for repeated manual intervention to set it. The purpose of this mechanism is to stably implement the aforementioned industrial data management methods in the real-world scenario of UHV power transmission disaster prevention, thereby achieving engineering consistency between dispatch strategies, energy supply constraints, and grid protection boundaries.

[0032] In a preferred embodiment of the present invention, the data processing center is further configured to collect the update node running status data of the edge sensing node after executing the dynamic scheduling instruction after the dynamic scheduling instruction is issued; The data processing center is also used to requantize the network vulnerability entropy state value based on the updated node running status data, and use the requantized network vulnerability entropy state value as a feedback label to iteratively update the internal weight parameters of the lightweight measurement model in order to achieve robust control optimization.

[0033] This embodiment provides a robust optimization mechanism for recovering the effects of scheduling; specifically, in the aforementioned implementation, the lightweight measurement model is already able to quantify the network vulnerability based on the node's running status and drive scheduling decisions. However, the disaster disturbances, equipment aging, node power supply differences, and local wireless environment of the transmission corridor may vary with the seasons and line operating conditions; If the model parameters are fixed for a long time, deviations may occur in new scenarios, such as reacting too slowly to a certain type of retransmission degradation or overreacting to a certain type of buffer fluctuation. Therefore, in this embodiment, after the dynamic scheduling command is issued, the running status data of the updated node after execution is collected, and the lightweight calculation model is corrected accordingly. Specifically, the so-called feedback labels are not manually labeled category results, but rather the results of re-observation of the actual network state after the scheduling actions are implemented; If the update status shows a decrease in cache, a reduction in retransmission, and manageable power supply pressure after the system issues a regular polling command, it indicates that the previous vulnerability assessment and scheduling match were good. If the update status continues to deteriorate after the system issues a degradation command, or even if the cache of the core node continues to rise, it indicates that the original model may have underestimated the risk or given insufficient sensitivity to certain types of instability signals. The data processing center uses the requantized network vulnerability entropy state value as a feedback label to iteratively update the internal weight parameters of the lightweight calculation model, namely the feature weighting coefficients used to adjust the proportion of importance of different types of node operating state data before model normalization. This allows the influence of signals from the buffer, power supply, and retransmission dimensions in the final vulnerability entropy calculation result to be adaptively adjusted, thereby enabling the model to gradually approach the real network evolution law under the line, season, and power supply conditions. In a simplified analysis model, it is assumed that the pre-scheduling model pays more attention to retransmission status and less attention to battery health degradation; during a disaster, the system judges the risk to be acceptable based on this model, so it only issues regular polling commands. After one cycle, the update node running status showed that the cache did not decrease. On the contrary, due to insufficient power, multiple nodes failed to wake up. The fragile state after requantization was significantly higher than expected. The data processing center therefore concluded that the impact of power supply health factors on overall vulnerability under the current conditions of continuous low temperature icing on the line was underestimated, and thus increased the relevant weight in subsequent updates. After multiple rounds of feedback, the model will be better adapted to this type of disaster environment; conversely, if a certain road section is in a stable wind but occasionally has electromagnetic interference, the short-term increase in retransmission will not have a lasting impact on the final survivability, and the corresponding weight can be adjusted back appropriately. Regarding anomaly handling mechanisms, model updates should not be over-adjusted on a single abnormal sample to avoid solidifying occasional noise into long-term patterns. Therefore, a minimum sample size, an upper limit for the update step size, or updates by line partition can be set. If the feedback data after scheduling is severely lacking, the parameter iteration of this round should be paused, and the current model should continue to be used. If the system is at the highest level of emergency response, priority should be given to ensuring scheduling stability. At this time, the model only records samples and does not perform large-scale online updates. The model will be uniformly verified offline after the disaster. If some nodes cause long-term distortion of feedback status due to hardware damage, they should be removed from the model learning samples to prevent contamination of the weight adjustment direction. During the aforementioned mountain corridor disaster, the system implemented deep hibernation for non-core nodes and maintained high-priority monitoring for core nodes. Within one cycle, the data processing center collected new node operating statuses and found that the retransmission rate of core nodes had decreased significantly. However, due to the continuous low temperature and icing that blocked light, the power supply recovery of some nodes was still not ideal. The requantified network vulnerability status indicates that while this scheduling did alleviate bandwidth congestion, the risk in the power supply dimension remains high. Therefore, the central system will appropriately enhance the sensitivity of power supply attenuation-related calculations in subsequent iterations, so that the next round of scheduling can identify power preservation needs earlier. The purpose of this mechanism is to enable the network vulnerability assessment model to move beyond a one-time setting and instead continuously correct itself based on real feedback after scheduling, thereby achieving robust control optimization that better fits the actual conditions on site.

[0034] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A power transmission disaster prevention monitoring system for industrial data management, characterized in that, The power transmission disaster prevention and monitoring system includes: Multiple edge sensing nodes distributed in the target area are networked to form a sensing network. The edge sensing nodes are used to collect industrial status data and node operation status data of the target area. An edge computing gateway, which is communicatively connected to the edge sensing node and has adjustable transmission bandwidth; A data processing hub, which is communicatively connected to the edge sensing node through the edge computing gateway; The data processing center is used to receive the industrial status data and the node operation status data, and quantitatively calculate the network vulnerability entropy state value of the sensing network based on the node operation status data. The data processing center is also used to calculate the trend of change by combining the stored historical network vulnerability entropy state value sequence, and predict the countdown rate of the sensing network approaching the preset failure time threshold based on the network vulnerability entropy state value and the trend of change, and generate dynamic scheduling instructions according to the countdown rate and the network vulnerability entropy state value. The data processing center is also used to send the dynamic scheduling instructions to the edge sensing nodes to comprehensively manage the data transmission and computing power status of the edge sensing nodes; wherein, the dynamic scheduling instructions include at least regular polling instructions and suboptimal degradation scheduling instructions.

2. The power transmission disaster prevention monitoring system for industrial data management according to claim 1, characterized in that, The node operating status data includes at least cache queue occupancy data, battery health decay data, and communication data packet retransmission rate data. The data processing center stores a lightweight calculation model; The data processing center is also used to input the cache queue occupancy rate data, the battery health decay data, and the communication data packet retransmission rate data into the lightweight calculation model, and use the lightweight calculation model to output the network vulnerability entropy state value. The lightweight measurement model is configured to perform feature fusion and uncertainty quantification on the cache queue occupancy rate data, the battery health decay data, and the communication data packet retransmission rate data to obtain the network vulnerability entropy state value.

3. The power transmission disaster prevention monitoring system for industrial data management according to claim 1, characterized in that, The data processing center is also used to compare the network vulnerability entropy state value with a preset fatal threshold: When the network vulnerability entropy state value is greater than or equal to the preset fatal threshold, the data processing center generates the suboptimal degradation scheduling instruction; When the network vulnerability entropy state value is less than the preset fatal threshold, the data processing center generates the regular polling instruction; The conventional polling instruction is used to instruct the edge sensing node to perform data collection and transmission according to a preset sleep cycle.

4. The power transmission disaster prevention monitoring system for industrial data management according to claim 3, characterized in that, The edge sensing nodes are divided into core nodes and non-core nodes; The suboptimal degradation scheduling instructions include at least data discard instructions and communication cutoff instructions; The data processing center, after generating the suboptimal degradation scheduling instruction, sends the data discard instruction and the communication cut-off instruction to the non-core node, instructing the non-core node to stop sending the industrial status data it collects, i.e. non-core area data, and enter a deep sleep state where its data acquisition module is shut down and only its wake-up monitoring circuit is powered, thereby releasing communication bandwidth for the core node and preserving the core node's power.

5. The power transmission disaster prevention monitoring system for industrial data management according to claim 4, characterized in that, The data processing center has a pre-stored generative data inference model, which is a neural network model trained based on historical industrial status data and used to establish the spatiotemporal mapping relationship between core area data and non-core area data. The data processing center is also used to extract the industrial status data, i.e. core area data, that is, after issuing the suboptimal degraded scheduling instruction to the non-core node; The data processing center is also used to input the core area data into the generative data inference model, generate inference filling data based on the spatiotemporal mapping relationship using the generative data inference model, and reconstruct the non-core area data using the inference filling data to maintain the global integrity of the industrial status data.

6. The power transmission disaster prevention monitoring system for industrial data management according to claim 1, characterized in that, The industrial status data includes sudden anomaly alarm data; The data processing center is also used to extract the data features of the sudden abnormal alarm data when the generation rate of the sudden abnormal alarm data is detected to be greater than the preset alarm storm threshold, and to identify the authenticity of the sudden abnormal alarm data based on the deviation of the data features from the pre-established normal fluctuation range of the device status: when the deviation is greater than the preset abnormal threshold, it is determined to be a valid alarm data of a true alarm; otherwise, it is determined to be redundant alarm data of a false alarm. The data processing center is also used to filter the redundant alarm data and reallocate the transmission bandwidth of the edge computing gateway based on the filtered valid alarm data.

7. The power transmission disaster prevention monitoring system for industrial data management according to claim 3, characterized in that, The preset failure time threshold is the maximum time span during which the system continuously fails to receive valid data. The specific method by which the data processing center calculates the trend of change and predicts the countdown rate is as follows: based on the time series prediction algorithm, the slope change of the historical network vulnerability entropy state value sequence is output, and the countdown rate is generated in combination with the preset failure time threshold.

8. A power transmission disaster prevention monitoring system for industrial data management according to any one of claims 1 to 6, characterized in that, The target area is the ultra-high voltage power transmission corridor area, and the industrial status data is micro-meteorological composite disaster data; The preset failure time threshold is a fixed time threshold pre-stored in the data processing center. This time threshold is calibrated according to the preventive power outage triggering time of the power grid-level security defense system. The edge sensing node relies on wind and solar power harvesting devices to obtain electricity.

9. A power transmission disaster prevention monitoring system for industrial data management according to claim 2, characterized in that, The data processing center is also used to collect the update node running status data of the edge sensing node after executing the dynamic scheduling instruction after the dynamic scheduling instruction is issued; The data processing center is also used to requantize the network vulnerability entropy state value based on the running status data of the update node, and use the requantized network vulnerability entropy state value as a feedback label to iteratively update the internal weight parameters of the lightweight calculation model in order to achieve robust control optimization.

Images