Power operation data monitoring method and system based on cloud network

Abstract

The application relates to the technical field of power monitoring, and discloses a power operation data monitoring method and system based on a cloud network, the method comprising the following steps: acquiring power equipment operation state data flow, uploading the operation state data flow to an edge intelligent gateway; the edge intelligent gateway acquires an abnormality index determined according to the operation state data flow, and uploads the operation state data flow to a corresponding cloud network resource pool in stages according to the abnormality index; after the cloud network resource pool receives the operation state data flow, the cloud network resource pool is elastically expanded according to the abnormality index. Through the method and the corresponding system, the matching between resources and data importance can be ensured, the insufficient computing power can be avoided, and the cloud network resource pool can be dynamically allocated on demand to improve resource utilization.

Description

Technical Field

[0001] This invention relates to the field of power monitoring technology, and in particular to a method and system for monitoring power operation data based on cloud networks. Background Technology

[0002] As power systems evolve towards large-scale smart grids, the number of power devices in the grid is surging, leading to an exponential increase in the total amount of power operation status data. Against this backdrop, existing data monitoring solutions suffer from the following prominent shortcomings:

[0003] First, at the data acquisition and transmission level, existing methods typically employ a uniform data transmission strategy; that is, operational status data generated by various power equipment (such as main transformers, transmission lines, distribution terminals, and smart meters) are all uploaded in parallel to the edge smart gateway or cloud network with the same transmission priority, rate, and channel. This one-size-fits-all data upload mechanism causes low-time-sensitivity business data (such as periodic metering data from smart meters) to compete for limited communication bandwidth resources with high-real-time-sensitivity critical data (such as line short-circuit fault alarm signals). The direct consequence is that critical data suffers significant transmission delays due to channel congestion, and may even experience packet loss, seriously threatening the power grid's real-time safety monitoring and rapid fault response capabilities.

[0004] Secondly, at the cloud platform data processing and resource management level, in order to cope with the surge in data processing demand, existing systems usually need to reserve a large amount of computing and storage resources based on peak load. However, the power operation data traffic has significant periodicity and volatility, resulting in a large amount of idle cloud service resources during off-peak periods; this extensive resource allocation model keeps cloud service costs high and causes waste. Summary of the Invention

[0005] Therefore, the purpose of this invention is to overcome the problems in the prior art where monitoring power operation data cannot be categorized and processed before uploading to the cloud network, resulting in the need to process a large amount of power operation data simultaneously and the inability to process abnormal data in a timely manner, and the problem that cloud network resources reserve a large amount of computing and storage resources based on peak load, but cannot be automatically expanded according to data, resulting in a waste of computing and storage resources. The invention provides a power operation data monitoring method and system based on a cloud network, which realizes hierarchical data uploading by calculating the abnormality index of the data, and expands the cloud network resource pool according to the degree of data abnormality after hierarchical uploading.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides a power operation data monitoring method based on a cloud network. The method includes: acquiring a power equipment operation status data stream and uploading the operation status data stream to an edge intelligent gateway; the edge intelligent gateway acquiring an anomaly index determined based on the operation status data stream and uploading the operation status data stream to corresponding cloud network resource pools in a tiered manner according to the anomaly index; and the cloud network resource pool receiving the operation status data stream and elastically expanding it according to the anomaly index.

[0007] Preferably, the edge intelligent gateway obtains the anomaly index based on the operating status data stream, including: extracting historical operating status data of the power equipment based on timestamps; calculating periodic anomaly values ​​of the triplet data stream based on the historical operating status data, wherein the triplet data stream is: device ID-timestamp-monitoring value; obtaining historical data of the power equipment operating status data within a preset time window; obtaining a predicted value through the anomaly prediction model of the edge intelligent gateway; obtaining a predicted anomaly value by subtracting the monitored value from the predicted value; and obtaining the anomaly index of the operating status data stream by weighted summation of the periodic anomaly value and the predicted anomaly value.

[0008] Preferably, before obtaining the anomaly index of the operating status data stream, the method further includes: uniformly converting the private protocol operating status data streams of different power devices into a standard format supported by the edge intelligent gateway to generate the triplet data stream; cleaning the triplet data stream, wherein the cleaning includes: filling in missing values ​​of monitoring values ​​in the triplet data stream through linear interpolation; replacing instantaneous jump values ​​of the monitoring values ​​with the average of valid monitoring values ​​within a preset time window; and replacing invalid values ​​of the monitoring values ​​with the previous valid monitoring value.

[0009] Preferably, the process of uploading the operational status data stream to the corresponding cloud network resource pool in a tiered manner according to the anomaly index includes: adding tiered tags to the operational status data stream based on the anomaly index, wherein the tiered tags include low, medium, and high; maintaining three sending buffer queues at the edge intelligent gateway, wherein the three sending buffer queues include: a low-priority data buffer queue, a medium-priority data buffer queue, and a high-priority data buffer queue; the low-priority data buffer queue caches operational status data with low tiered tags, the medium-priority data buffer queue caches operational status data with medium tiered tags, and the high-priority data buffer queue caches operational status data with high tiered tags.

[0010] Preferably, the method further includes: the low-priority data buffer queue uploading its running status data stream to the corresponding cloud network resource pool in batches via a first triggering method; the medium-priority data buffer queue uploading its running status data stream to the corresponding cloud network resource pool in batches via a second triggering method; and the high-priority data buffer queue uploading its running status data stream to the corresponding cloud network resource pool in batches via a third triggering method; wherein the triggering frequency of the first triggering method is lower than the triggering frequency of the second triggering method, and the triggering frequency of the second triggering method is lower than the triggering frequency of the third triggering method.

[0011] Preferably, after receiving the operational status data stream, the cloud network resource pool elastically expands according to the anomaly index, including: storing the data from the high-priority data buffer queue into the corresponding regional anomaly index pool of the cloud network resource pool; mapping the abnormal devices corresponding to the data through the triple data stream; setting clustering parameters, including distance threshold and density threshold, calling the DBSCAN clustering algorithm, and performing cluster analysis on the abnormal devices in the regional anomaly index pool according to the clustering parameters; if the clustering identifies an abnormal cluster, then marking the cloud network resource pool corresponding to the regional anomaly index pool as an object to be expanded, and determining the number of instances to be expanded in the cloud network resource pool according to the size of the abnormal cluster.

[0012] Preferably, the number of expanded instances is determined based on the size of the anomalous cluster, including: the number of expanded instances is positively correlated with the number of devices in the anomalous cluster.

[0013] Preferably, it further includes: if the abnormal cluster is identified as having dissipated, a reverse operation is triggered, wherein the reverse operation is used to reduce the number of instances in the cloud network resource pool down to a first preset number.

[0014] Preferably, the method further includes: analyzing the abnormal event sequence through the cloud network for data in the high-priority data buffer queue, instantly determining the fault type, and issuing preliminary control commands to the edge smart gateway; and optimizing the anomaly prediction model of the edge smart gateway based on the data in the medium-priority and low-priority data buffer queues.

[0015] Secondly, to address the aforementioned technical problems, this invention also proposes a cloud network-based power operation data monitoring system. The system includes: an edge intelligent gateway module for acquiring power equipment operation status data streams and uploading the operation status data streams to the edge intelligent gateway; an anomaly index determination module, where the edge intelligent gateway acquires an anomaly index based on the operation status data streams and uploads the operation status data streams to corresponding cloud network resource pools in tiers according to the anomaly index; and an elastic expansion module, where the cloud network resource pool receives the operation status data streams and elastically expands according to the anomaly index.

[0016] Compared with the prior art, the above-described technical solution of the present invention has the following advantages:

[0017] The cloud network-based power operation data monitoring method described in this invention calculates the data anomaly index to achieve hierarchical data uploading, and expands the cloud network resource pool according to the degree of data anomaly after hierarchical uploading.

[0018] This involves uploading power operation data streams to an edge smart gateway to leverage the proximity of edge nodes, reducing data transmission latency, and preventing network congestion caused by massive amounts of raw data flowing directly into the cloud.

[0019] Based on a preset algorithm, the received data stream is calculated for an anomaly index and uploaded in a tiered manner to achieve on-demand transmission, reduce the network and cloud resources occupied by invalid data, ensure that resources and data importance are matched, and avoid insufficient computing power.

[0020] The elastic expansion mechanism based on the anomaly index enables the cloud network resource pool to be dynamically allocated on demand, shrinking resources under low load and precisely expanding under high load, thereby improving resource utilization, avoiding idle waste caused by traditional fixed resource configuration, and reducing the operation and maintenance costs of cloud infrastructure. Attached Figure Description

[0021] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein...

[0022] Figure 1 This is a flowchart of a cloud network-based power operation data monitoring method in a preferred embodiment of the present invention;

[0023] Figure 2 This is a structural block diagram of a cloud-based power operation data monitoring system in a preferred embodiment of the present invention. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0025] Example 1: As Figure 1 As shown in the figure, this invention discloses a power operation data monitoring method based on a cloud network. The method includes: acquiring a power equipment operation status data stream and uploading the operation status data stream to an edge intelligent gateway; the edge intelligent gateway acquiring an anomaly index determined based on the operation status data stream and uploading the operation status data stream to the corresponding cloud network resource pool according to the anomaly index; and the cloud network resource pool receiving the operation status data stream and elastically expanding according to the anomaly index.

[0026] In specific application scenarios, power equipment collects real-time operational status data streams through intelligent terminals such as smart sensors. This power equipment covers the entire power generation, transmission, transformation, distribution, and consumption chain, including thermal power units in the power generation stage, high-voltage overhead lines in the transmission stage, and main transformers in the transformation stage. The operational status data stream includes current, and this data stream is uploaded to a locally deployed edge smart gateway. Leveraging the proximity of edge nodes reduces data transmission latency and avoids network congestion caused by massive amounts of raw data flooding the cloud. The edge smart gateway calculates an anomaly index on the received data stream based on a preset algorithm. Based on the anomaly index's classification rules, the edge smart gateway selectively uploads data: data with low classification labels is uploaded with low-frequency summary information, data with medium classification labels is uploaded with medium-frequency data, and data with high classification labels is uploaded with high-frequency data, achieving on-demand data upload. To reduce network and cloud resource consumption by invalid data transmission, the edge-side hierarchical upload mechanism minimizes invalid data transmission. Only high-risk data is uploaded in full and at high frequency, avoiding congestion of the backbone network by massive amounts of raw data. At the same time, edge preprocessing shortens the response link from data collection to analysis, reducing the initial judgment latency of critical anomaly data from seconds to milliseconds. After receiving data streams of corresponding levels, cloud network resource pools of different levels dynamically adjust resource configuration based on the anomaly index for data with high level labels. Through dynamic scheduling in cloud-edge collaboration, resource matching with data importance is ensured, avoiding resource waste or insufficient computing power. The elastic expansion mechanism based on the anomaly index enables cloud network resource pools to dynamically allocate resources on demand, shrinking resources under low load and precisely expanding under high load, improving resource utilization, avoiding idle waste caused by traditional fixed resource configuration, and reducing the operation and maintenance costs of cloud infrastructure.

[0027] In one embodiment of the present invention, the edge intelligent gateway obtains an anomaly index based on the operating status data stream, comprising: extracting historical operating status data of the power equipment based on a timestamp; calculating periodic anomaly values ​​of a triplet data stream based on the historical operating status data, wherein the triplet data stream is: device ID-timestamp-monitoring value; obtaining historical data of the power equipment operating status data within a preset time window; obtaining a predicted value through the anomaly prediction model of the edge intelligent gateway; obtaining a predicted anomaly value by subtracting the monitored value from the predicted value; and obtaining the anomaly index of the operating status data stream by weighted summation of the periodic anomaly value and the predicted anomaly value.

[0028] In specific application scenarios, the monitored values ​​include electrical parameters and physical state parameters. The electrical parameters directly reflect the basic electrical performance of the power equipment. These electrical parameters include three-phase current, three-phase voltage, zero-sequence current, zero-sequence voltage, and voltage imbalance. In this embodiment, zero-sequence current is used as an example. The physical state parameters include winding temperature, oil temperature of oil-immersed transformers, circuit breaker contact temperature, cable joint temperature, surface temperature of distribution cabinet, and generator vibration frequency. Here, an edge intelligent gateway continuously receives data streams from off-duty transformer monitoring terminals, forming a triplet data stream, such as {Device ID:DT-City A-001, Timestamp:2023-07-26 14:00:00, Monitoring Value: 285A}, and calculates the periodicity. Outlier detection aims to identify whether the device's operating status deviates from its historical patterns. Based on the current data's timestamp (Wednesday 14:00), the edge smart gateway extracts the current values ​​at 14:00 for the previous four consecutive Wednesdays from its locally cached historical database: July 19, July 12, July 5, and June 28. The historical current values ​​for these four periods are [275A, 268A, 270A, 272A]. The edge smart gateway calculates the average (μ_hist) and standard deviation (σ_hist) of this set of historical data: μ_hist = (275 + 268 + 270 + 272) / 4 = 271.25A, σ_hist = calculated standard deviation (...) ≈ 2.86A.

[0029] The Z-Score standardization method is used to calculate the deviation of the current monitoring value from the historical data of the same period, i.e., the periodic outlier (Pe). Then, the Z-Score standardization method is used to calculate the deviation of the current monitoring value from the historical data of the same period, i.e., the periodic outlier (Pe), Pe=|(current monitoring value-μ_hist) / σ_hist|. Substituting into the calculation: Pe=|(285-271.25) / 2.86|≈4.82. This value Pe=4.82 is much greater than 2, indicating that the current current is extremely significantly abnormally high compared with the normal level of the same period in history. This means that the current load of the transformer is likely abnormal. The calculation of the periodic outlier takes into account the periodic pattern and can detect chronic problems caused by equipment aging and slow load increase.

[0030] The calculation and prediction of outliers aims to predict the expected value of a device at the next moment based on its recent operating trends, thereby detecting sudden anomalies. The edge smart gateway obtains the current data sequence within a preset time window (e.g., the past 30 minutes) preceding the current moment from the real-time data stream as input to the model. The edge smart gateway has a built-in lightweight time-series prediction model, such as a simplified model based on ARIMA (Autoregressive Integral Moving Average) or one-dimensional convolution. This model takes the data sequence from the previous 30 minutes as input and outputs a predicted current value for the next moment (i.e., 14:00). Calculating short-term trends can detect acute problems caused by lightning strikes, short circuits, etc., significantly reducing the false negative rate.

[0031] Subsequently, the predicted outlier (Pf) is calculated, which is the absolute difference between the actual monitored value and the predicted value. A weighted summation is then performed to obtain a comprehensive anomaly index, which combines Pe (reflecting long-term deviations from patterns) and Pf (reflecting short-term trend changes) into a single comprehensive index. To ensure that Pe and Pf can be added together, they are first normalized using Min-Max, mapping them to the range of 0-100. The alarm threshold for Pe is set to 3. Therefore, Pe_normalized = min((Pe / 3)×100, 100) = min((4.82 / 3)×100, 100) ≈ 10 0, Pf_normalized=min((Pf / 15)×100,100)=min((19 / 15)×100,100)≈100, assign appropriate weights to the two indicators (the sum of the weights is 1), calculate the comprehensive anomaly index AI, AI=α×Pe_normalized+β×Pf_normalized, in this embodiment the weights can be set α=0.4, β=0.6, substitute into the calculation: AI=0.4×100+0.6×100=100; Pe_normalized represents the periodic outlier after normalization, Pf_normalized represents the predicted outlier after normalization. The core definition is that when the edge intelligent gateway receives an extremely high anomaly index of AI=100, according to preset rules (AI > first preset threshold is high priority data stream), the edge intelligent gateway immediately marks the data of this transformer as the highest priority (alarm data stream). The edge intelligent gateway activates the emergency upload channel, uploading the full real-time data stream and cached data to the emergency command resource pool in the cloud, and is likely to trigger local audible and visual alarms simultaneously, notifying inspection personnel to handle the situation urgently. Periodic analysis (Pe) can issue early warnings as soon as the device status begins to deviate from its historical normal trajectory, achieving early warning. The entire calculation process is completed only on the edge side; the algorithm is lightweight but effective, making intelligent decisions without cloud intervention, fully demonstrating the value of edge intelligence.

[0032] In one embodiment of the present invention, before obtaining the anomaly index of the operating status data stream, the method further includes: uniformly converting the private protocol operating status data streams of different power devices into a standard format supported by the edge intelligent gateway to generate the triplet data stream; cleaning the triplet data stream, wherein the cleaning includes: completing missing values ​​of monitoring values ​​in the triplet data stream through linear interpolation; replacing instantaneous jump values ​​of the monitoring values ​​with the average of valid monitoring values ​​within a preset time window; and replacing invalid values ​​of the monitoring values ​​with the previous valid monitoring value.

[0033] In specific application scenarios, such as the power Internet of Things (IoT), the devices accessed at the edge are from numerous manufacturers and come in a wide variety of models, resulting in diverse data access protocols (such as proprietary or industry protocols like Modbus, IEC104, and DLT645). The raw data stream often contains missing values, transient jumps, and invalid values ​​caused by transmission interference, sensor malfunctions, or data acquisition unit restarts. Directly using such "dirty data" to calculate the anomaly index can lead to false alarms and missed alarms, severely impacting the monitoring system's judgment. Therefore, this embodiment designs a standardized data preprocessing and cleaning process before calculating the anomaly index to ensure the consistency and high quality of the input data. Heterogeneous proprietary protocol data is converted into a unified, easily processed standard data format within the edge smart gateway. The edge smart gateway has pre-built parsing drivers for various protocols (such as Modbus TCP driver and IEC-104 driver). When the edge smart gateway receives raw data packets from different devices, it calls the corresponding driver for parsing. After parsing, the key device ID, timestamp, and monitoring value are extracted and encapsulated into a unified standard JSON format, generating a triplet data stream. Through protocol conversion, seamless access to power equipment from a wide range of sources can be achieved, solving the core pain point of equipment heterogeneity. After generating a standard triplet data stream, the edge smart gateway initiates a real-time cleaning pipeline.

[0034] Missing value completion: Due to network jitter, the edge smart gateway received data at 10:00:30, but the data packet at 10:01:00 was lost, and the next data was not received until 10:01:30; the cleaning pipeline detected the missing data point at 10:01:00. It uses linear interpolation to complete the data using the two valid data points before and after it; the previous point (10:00:30): current=271.0A, the next point (10:01:30): current=274.0A, the completion calculation: completed value = 271.0 + (274.0 - 271.0) × ((60 - 30) / (90 - 30)) = 271.0 + 3.0 × 0.5 = 272.5A.

[0035] For handling instantaneous jump values, if a current that normally fluctuates between 250A and 260A suddenly experiences a spike of 500A lasting only one sampling cycle, this is likely caused by electromagnetic interference. A dynamic threshold is set (e.g., the difference from the previous value exceeds 5 times the average fluctuation rate). When |500-255|=245A significantly exceeds the threshold, it is determined to be an instantaneous jump. Using this jump point as the center, valid monitoring values ​​(i.e., non-jumping, non-invalid values) within a preset time window (e.g., 2 points before and after, for a total of 5 points) are taken. The data within the window is: [253, 255, 500 (jump), 257, 256]. The average of these valid values ​​is calculated as: (253+255+257+256) / 4=255.25A. This average value of 255.25A is used to replace the original jump value of 500A.

[0036] Invalid value handling: Upon power-off and restart, the sensor may upload a NULL, NaN, or an illegal value far exceeding its range (such as -1 or 9999). Invalid values ​​are determined using a predefined rule base. Rules include whether the value is NULL or NaN, and whether the value exceeds the physical measurement range (e.g., current less than 0A or greater than 1000A). Once determined to be invalid, the system replaces it with the previous valid monitoring value. For example, in a data sequence of [255, 257, -1 (invalid), 259, ...], the edge smart gateway replaces the invalid value -1 with its previous valid value, 257A. High-quality data cleaning effectively removes noise and anomalies, providing a clean data source for subsequent anomaly index calculations and significantly reducing the false positive rate of the core algorithm.

[0037] In one embodiment of the present invention, the operation status data stream is uploaded to the corresponding cloud network resource pool in a hierarchical manner according to the anomaly index, including: adding hierarchical tags to the operation status data stream according to the anomaly index, the hierarchical tags including low, medium and high; maintaining three sending buffer queues at the edge intelligent gateway, the three sending buffer queues including: a low-priority data buffer queue, a medium-priority data buffer queue and a high-priority data buffer queue; the low-priority data buffer queue caches operation status data with low hierarchical tags, the medium-priority data buffer queue caches operation status data with medium hierarchical tags, and the high-priority data buffer queue caches operation status data with high hierarchical tags.

[0038] In specific application scenarios, the edge smart gateway maintains a label mapping table internally, which maps the calculated anomaly index value to three graded labels: low, medium, and high. When 0 < anomaly index <= the first preset threshold, the graded label is low; when the first preset threshold < anomaly index <= the second preset threshold, the graded label is medium; and when the anomaly index > the second preset threshold, the graded label is high.

[0039] In one embodiment of the present invention, the low-priority data buffer queue uploads the running state data stream therein to the corresponding cloud network resource pool in batches through the first triggering method; the medium-priority data buffer queue uploads the running state data stream therein to the corresponding cloud network resource pool in batches through the second triggering method; the high-priority data buffer queue uploads the running state data stream therein to the corresponding cloud network resource pool in batches through the third triggering method; wherein, the triggering frequency of the first triggering method is lower than that of the second triggering method, and the triggering frequency of the second triggering method is lower than that of the third triggering method.

[0040] In a specific application scenario, the first triggering method adopts a timing triggering mechanism, sets a relatively long time interval T1, and when the timer reaches T1, uploads the accumulated data in the queue to the historical data storage resource pool in the cloud network resource pool in batches; the second triggering method adopts a double-triggering mechanism of timing triggering and queue depth, sets a medium time interval T2 (T2 < T1), and at the same time sets a queue depth threshold D; when the timer reaches T2 or the data volume in the queue reaches the threshold D, immediately trigger the data upload to the stream processing resource pool in the cloud network resource pool; the third triggering method adopts an immediate triggering mechanism. Once data enters the high-priority queue, immediately start the upload process, and at the same time set a very short time interval T3 (T3 < T2) as a polling guarantee; the data will be directly uploaded to the real-time analysis and emergency resource pool in the cloud network resource pool. Through the above hierarchical triggering mechanism, the low-priority data is uploaded at a lower frequency, saving network bandwidth; the medium-priority data takes into account the system load while ensuring a certain degree of real-time performance; the high-priority data obtains an immediate response, ensuring rapid handling in case of emergencies. This multi-level triggering architecture effectively balances the utilization of system resources and the real-time requirements of services, providing an optimized data transmission solution for power operation data monitoring.

[0041] In one embodiment of the present invention, after the cloud network resource pool receives the running state data stream, it expands elastically according to the anomaly index, including: storing the data in the high-priority data buffer queue into the regional anomaly index pool of the corresponding cloud network resource pool; mapping the abnormal device corresponding to the data through the triple data stream; setting clustering parameters, including a distance threshold and a density threshold, calling the DBSCAN clustering algorithm, and performing clustering analysis on the abnormal devices in the regional anomaly index pool according to the clustering parameters; if an abnormal aggregation cluster is identified by clustering, mark the cloud network resource pool corresponding to the regional anomaly index pool as an object to be expanded, and determine the number of instances for expanding the cloud network resource pool according to the scale of the abnormal aggregation cluster.

[0042] In specific application scenarios, such as power monitoring, when multiple geographically adjacent devices malfunction almost simultaneously, it may indicate an impending cascading failure on a power line or in a particular area. This type of regional risk necessitates immediate deployment of significant cloud-based computing resources for collaborative analysis and simulation. The core of this embodiment lies in proactively predicting computing resource demands by identifying the clustering effect of malfunctioning devices in physical space, thereby achieving intelligent management from business risk to IT resource allocation.

[0043] The system continuously receives alarm data streams uploaded from the edge intelligent gateway via a high-priority data buffer queue, parses these data, and stores them in a temporary high-speed storage area called the Regional Anomaly Index Pool. The system maps the corresponding abnormal devices through the triplet data stream. Based on the device ID in the triplet, the system retrieves the latitude and longitude coordinates and geographical information such as the region of the device from a preset device asset database. Clustering parameters are set, for example, a distance threshold of 5 kilometers, indicating that abnormal devices within a 5-kilometer radius may be related; a density threshold of 3, indicating that at least 3 abnormal devices are required to form an "anomaly cluster." The analysis service within the resource pool is triggered periodically (e.g., every minute) or when the number of devices in the Regional Anomaly Index Pool reaches a certain threshold. It calls the DBSCAN clustering algorithm, using the latitude and longitude coordinates of all abnormal devices as input features. DBSCAN can effectively discover clusters of arbitrary shapes and identify discrete, independent anomalies (noise points), making it very suitable for the geographical distribution characteristics of power grid equipment. Assume there are 10 anomalous devices in the current regional anomaly index pool, with the following geographical distribution: Cluster A: Devices {DT-001, DT-002, DT-003, DT-004} are densely distributed in an industrial park (within a 3 km diameter). Cluster B: Devices {CB-101, CB-102, CB-103} are densely distributed in a residential area (within a 2 km diameter). Noise points: Devices {TR-055, TR-076, LN-210} are isolated, with a distance of more than 5 km between them. DBSCAN will identify two clusters: Cluster 1 (Cluster A), containing 4 devices; Cluster 2 (Cluster B), containing 3 devices; the remaining 3 devices are marked as noise.

[0044] After the clustering analysis is completed, all identified clusters are checked; if at least one cluster exists (i.e., a group of devices that are not noisy and meet the density conditions), the cloud network resource pool corresponding to the abnormal index pool in that area is immediately marked as an object that needs to be expanded.

[0045] In one embodiment of the present invention, determining the number of expanded instances based on the size of the anomalous cluster includes: the number of expanded instances is positively correlated with the number of devices in the anomalous cluster.

[0046] In specific application scenarios, the number of devices in the anomaly cluster directly determines the data processing requirements. The more anomaly devices there are, the larger the real-time data stream will be. Therefore, the number of expanded instances is positively correlated with the number of devices in the anomaly cluster to ensure that resource capacity matches the data volume and analysis complexity.

[0047] In one embodiment of the present invention, if it is identified that the abnormal cluster has dissipated, a reverse operation is triggered, wherein the reverse operation is used to reduce the number of instances in the cloud network resource pool down to a first preset number.

[0048] In specific application scenarios, the elastic scaling of cloud network resources not only includes scaling up based on demand but also necessarily scaling down when demand decreases. In power monitoring scenarios, if the large amount of computing resources temporarily allocated to a cluster of abnormal devices caused by a regional fault continues to run after it has been processed and dissipated, it will result in significant resource waste and increased costs. The first preset number refers to the baseline number of instances that the cloud network resource pool should maintain when there are no abnormal processing tasks. Periodically, health checks are performed on all currently marked abnormal clusters. In the latest DBSCAN clustering analysis, due to the improvement of device status, the original clusters no longer meet the clustering parameters (distance threshold and density threshold) spatially, causing the cluster to disappear from the clustering results. Once the system determines that one or all abnormal clusters have dissipated, the scaling down process is immediately triggered. The ultimate goal of scaling down is to reduce the number of instances in the cloud network resource pool from the current high level to the first preset number. By automatically and promptly scaling down, the idle waste of cloud computing resources during off-peak hours or after the fault subsides is avoided, significantly reducing operating costs. A complete elastic closed loop is formed. This method, combined with the expansion mechanism, constitutes a complete autonomous closed loop of demand perception, intelligent expansion, risk mitigation, and automatic scaling down, reflecting the intelligence of the system.

[0049] In one embodiment of the present invention, for data in the high-priority data buffer queue, the abnormal event sequence is analyzed through the cloud network to determine the fault type in real time and issue preliminary control instructions to the edge smart gateway; for data in the medium-priority data buffer queue and the low-priority data buffer queue, the anomaly prediction model of the edge smart gateway is optimized based on the data.

[0050] In specific application scenarios, when high-priority data streams enter the emergency command resource pool in the cloud, the complex event processing engine begins to work. It no longer views individual data points in isolation, but analyzes a series of abnormal data generated by a device or region within a short period as an abnormal event sequence. The received sequence is as follows: t0 time: {Device ID: TR-001, Current: Sudden Rise, Oil Temperature: Normal, Vibration: Normal}, t1 time (t0+2s): {Device ID: TR-001, Current: Maintains High Level, Oil Temperature: Begins to Rise Slowly, Vibration: High-Frequency Components Appear}, t2 time (t1+5s): {Device ID: TR-001, Current: Fluctuation, Oil Temperature: Accelerates Rise, Vibration: Increased Amplitude}. This sequence is then matched against a pre-stored fault feature library in the cloud. The sequence highly matches the fault characteristics of the initial stage of a winding short circuit (current first suddenly rises, then oil temperature rises due to internal heating, and vibration intensifies due to changes in electromagnetic force). A clear fault type diagnosis result is generated, such as fault type: winding short circuit. Path; Confidence level: 85%; Recommended operation: Immediately adjust the relay protection settings and recommend inspection; Send the instruction to the edge smart gateway that generated the data; After receiving the instruction, the edge smart gateway can execute it directly through its control module without complex calculations; The cloud network uses the massive amount of aggregated medium and low priority data to periodically retrain or optimize the anomaly prediction model used on the edge side. Low priority data is used to establish a more accurate benchmark for normal equipment operation, while medium priority data helps the model learn how anomalies begin and evolve. After training, the cloud network generates a lightweight model incremental update package. When the cloud network load is low (such as at night), the cloud network securely sends the update package to the edge smart gateway in the specific area.

[0051] Example 2: This embodiment of the invention provides a power operation data monitoring system based on a cloud network, referring to... Figure 2 As shown, the system includes: an edge intelligent gateway module for acquiring power equipment operation status data streams and uploading the operation status data streams to the edge intelligent gateway; an anomaly index determination module, wherein the edge intelligent gateway acquires an anomaly index based on the operation status data streams and uploads the operation status data streams to corresponding cloud network resource pools in a tiered manner according to the anomaly index; and an elastic expansion module, wherein the cloud network resource pool receives the operation status data streams and performs elastic expansion based on the anomaly index.

[0052] The cloud-based power operation data monitoring system described in this embodiment of the invention is used to implement the cloud-based power operation data monitoring method described in the above embodiments. Both are based on the same inventive concept and have the same technical effects, which will not be repeated here.

[0053] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0057] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A power operation data monitoring method based on cloud networks, characterized in that, The method includes: Acquire the power equipment operating status data stream and upload the operating status data stream to the edge smart gateway; The edge intelligent gateway obtains an anomaly index based on the operating status data stream, and uploads the operating status data stream to the corresponding cloud network resource pool in a hierarchical manner according to the anomaly index. After receiving the operational status data stream, the cloud network resource pool elastically expands according to the anomaly index. The edge intelligent gateway obtains an anomaly index based on the operating status data stream, including: extracting historical operating status data of the power equipment based on timestamps; calculating periodic anomaly values ​​of the triplet data stream based on the historical operating status data, wherein the triplet data stream is: device ID-timestamp-monitoring value; obtaining historical data of the power equipment operating status data within a preset time window; obtaining a predicted value through the anomaly prediction model of the edge intelligent gateway; obtaining a predicted anomaly value by subtracting the monitored value from the predicted value; and obtaining the anomaly index of the operating status data stream by weighted summation of the periodic anomaly value and the predicted anomaly value. After receiving the operational status data stream, the cloud network resource pool elastically expands based on the anomaly index, including: storing data from the high-priority data buffer queue into the corresponding regional anomaly index pool of the cloud network resource pool; mapping the abnormal devices corresponding to the data through the triple data stream; setting clustering parameters, including distance threshold and density threshold, and calling the DBSCAN clustering algorithm to perform cluster analysis on the abnormal devices in the regional anomaly index pool based on the clustering parameters; if the clustering identifies an abnormal cluster, the cloud network resource pool corresponding to the regional anomaly index pool is marked as an object to be expanded, and the number of instances to be expanded in the cloud network resource pool is determined based on the size of the abnormal cluster.

2. The power operation data monitoring method based on cloud network according to claim 1, characterized in that, Before obtaining the anomaly index of the running status data stream, the following steps are also included: The private protocol operation status data streams of different power devices are uniformly converted into the standard format supported by the edge smart gateway to generate the triple data stream. The triplet data stream is cleaned, which includes: filling in missing values ​​of the monitoring values ​​in the triplet data stream using linear interpolation; replacing instantaneous jump values ​​of the monitoring values ​​with the average of valid monitoring values ​​within a preset time window; and replacing invalid values ​​of the monitoring values ​​with the previous valid monitoring value.

3. The power operation data monitoring method based on cloud network according to claim 1, characterized in that, The operational status data stream is uploaded to the corresponding cloud network resource pool in a tiered manner based on the anomaly index, including: The operation status data stream is labeled with a classification tag based on the anomaly index, and the classification tag includes low, medium and high. The edge intelligent gateway maintains three transmission buffer queues, which include: a low-priority data buffer queue, a medium-priority data buffer queue, and a high-priority data buffer queue. The low-priority data buffer queue cache level label is low-level running status data, the medium-priority data buffer queue cache level label is medium-level running status data, and the high-priority data buffer queue cache level label is high-level running status data.

4. The power operation data monitoring method based on cloud network according to claim 3, characterized in that, Also includes: The low-priority data buffer queue uploads its running status data stream to the corresponding cloud network resource pool in batches through a first triggering method. The medium-priority data buffer queue uploads its running status data stream to the corresponding cloud network resource pool in batches through a second triggering method. The high-priority data buffer queue uploads its running status data stream to the corresponding cloud network resource pool in batches through a third triggering method. The triggering frequency of the first triggering method is lower than that of the second triggering method, and the triggering frequency of the second triggering method is lower than that of the third triggering method.

5. The power operation data monitoring method based on cloud network according to claim 1, characterized in that, The number of expanded instances is determined based on the size of the anomalous cluster, including: The number of expanded instances is positively correlated with the number of devices in the anomalous cluster.

6. The power operation data monitoring method based on cloud network according to claim 1, characterized in that, Also includes: If the abnormal cluster is identified as having dissipated, a reverse operation is triggered, which reduces the number of instances in the cloud network resource pool down to a first preset number.

7. The power operation data monitoring method based on cloud network according to claim 1, characterized in that, Also includes: For data in the high-priority data buffer queue, the abnormal event sequence is analyzed through the cloud network to determine the fault type in real time and issue preliminary control commands to the edge smart gateway. The anomaly prediction model of the edge smart gateway is optimized based on the data from the medium-priority data buffer queue and the low-priority data buffer queue.

8. A cloud-based power operation data monitoring system, characterized in that, The system includes: Upload to the edge smart gateway module to obtain the power equipment operation status data stream and upload the operation status data stream to the edge smart gateway; The anomaly index determination module involves the edge intelligent gateway acquiring an anomaly index determined based on the operating status data stream, and then uploading the operating status data stream to the corresponding cloud network resource pool in a tiered manner according to the anomaly index. The elastic expansion module allows the cloud network resource pool to elastically expand based on the anomaly index after receiving the running status data stream. The edge intelligent gateway obtains an anomaly index based on the operating status data stream, including: extracting historical operating status data of the power equipment based on timestamps; calculating periodic anomaly values ​​of the triplet data stream based on the historical operating status data, wherein the triplet data stream is: device ID-timestamp-monitoring value; obtaining historical data of the power equipment operating status data within a preset time window; obtaining a predicted value through the anomaly prediction model of the edge intelligent gateway; obtaining a predicted anomaly value by subtracting the monitored value from the predicted value; and obtaining the anomaly index of the operating status data stream by weighted summation of the periodic anomaly value and the predicted anomaly value. After receiving the operational status data stream, the cloud network resource pool elastically expands based on the anomaly index, including: storing data from the high-priority data buffer queue into the corresponding regional anomaly index pool of the cloud network resource pool; mapping the abnormal devices corresponding to the data through the triple data stream; setting clustering parameters, including distance threshold and density threshold, and calling the DBSCAN clustering algorithm to perform cluster analysis on the abnormal devices in the regional anomaly index pool based on the clustering parameters; if the clustering identifies an abnormal cluster, the cloud network resource pool corresponding to the regional anomaly index pool is marked as an object to be expanded, and the number of instances to be expanded in the cloud network resource pool is determined based on the size of the abnormal cluster.

Images