A modular intelligent box-type substation operation monitoring method and system

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a modular intelligent box-type substation operation monitoring method and system, comprising: collecting the operation data of the substation, and treating the multiple operation data collected at the same time as a high-dimensional data point; using the optimized LOF algorithm to calculate the abnormality score of each high-dimensional data point, and monitoring the operation status with the abnormality score, wherein the optimized LOF algorithm comprises: calculating the optimal K value of each high-dimensional data point according to the load change degree of each high-dimensional data point; wherein the load change degree of each high-dimensional data point is calculated. The present invention optimizes the K value of the LOF algorithm through cluster analysis and load change degree calculation, and dynamically adjusts the K value of the LOF algorithm, thereby solving the problem of inaccurate detection of the fixed K value in the traditional LOF algorithm when the load fluctuates, improving the sensitivity and accuracy of abnormality detection, and avoiding misjudgment caused by changes in the data fluctuation range.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a modular intelligent box-type substation operation monitoring method and system. Background Art

[0002] Traditional substation monitoring systems typically rely on manual inspections and fixed monitoring equipment for operational status monitoring. However, these methods have limitations. First, manual inspections rely on human efficiency and judgment, are susceptible to human factors, and cannot achieve real-time, comprehensive monitoring. Second, traditional monitoring equipment is mostly static, lacking flexibility and adaptability, and cannot intelligently adjust and optimize according to the changing needs of the power grid. Finally, insufficient centralized processing and analysis capabilities for monitoring data make it difficult to quickly and accurately locate the problem when a fault or anomaly occurs, impacting the reliability and stability of the power system. In recent years, modular substations have gradually become a new substation construction model. By modularizing and standardizing different electrical equipment and functions, modular substations improve the flexibility and scalability of substations in design, construction, operation, and maintenance. However, despite the many advantages of modular substations, traditional operational monitoring methods still suffer from the aforementioned issues, preventing them from fully realizing their advantages.

[0003] Currently, when monitoring substation operating status in real time, the LOF algorithm can be used to detect abnormal data during operation. The LOF algorithm is a density-based, unsupervised anomaly detection algorithm that identifies outliers by quantifying the relative difference between the local density of a data point and the density of its neighborhood. Its core assumption is that the local density of a normal point is similar to that of its neighborhood, while the local density of an outlier is significantly lower than that of its neighborhood. However, the distribution of data in various dimensions can vary significantly under varying substation loads. For example, under high load conditions, current amplitudes and temperatures increase, leading to rapid changes in data across various dimensions. Under low load conditions, the demand for these dimensions decreases significantly, resulting in less variation in data across these dimensions. Therefore, under different load conditions, the local density of data collected at different times varies. Traditional LOF algorithms, which use a fixed K value to detect anomalies in constantly changing data across various dimensions, produce inaccurate results. Summary of the Invention

[0004] The present invention provides a modular intelligent box-type substation operation monitoring method and system, aiming to solve the problem of insufficient adaptability to load fluctuations in the traditional LOF algorithm using a fixed K value in the related art.

[0005] In the first aspect, the present invention provides a modular intelligent box-type substation operation monitoring method, comprising: collecting the operation data of the substation, and treating the multiple operation data collected at the same time as a high-dimensional data point; using the optimized LOF algorithm to calculate the abnormality score of each high-dimensional data point, and monitoring the operation status with the abnormality score, wherein the optimized LOF algorithm includes: calculating the optimal K value of each high-dimensional data point according to the load change degree of each high-dimensional data point; wherein calculating the load change degree of each high-dimensional data point includes: clustering all high-dimensional data points to obtain multiple clusters, calculating the difference between any high-dimensional data point and the high-dimensional data points in the cluster where it is located; based on the product of the difference degree of any high-dimensional data point and the load change weight of the high-dimensional data point as the load change degree of the high-dimensional data point, wherein the load change weight of the high-dimensional data point is positively correlated with the collection time interval between the high-dimensional data point and all the high-dimensional data points in the cluster where it is located. The load change degree of the high-dimensional data point is calculated by cluster analysis, and the load change weight is determined in combination with the collection time interval. The dynamic change characteristics of the load during the operation of the substation are fully considered. For example, when the load fluctuates dramatically, data points may show greater differences, and the weights will be adjusted accordingly, making the anomaly score more reflective of the actual operating status. This dynamic monitoring capability is particularly suitable for substations with complex operating environments and frequently changing loads.

[0006] Furthermore, the degree of difference is calculated, including: calculating the Euclidean distance between any high-dimensional data point and the cluster center of its cluster, and the average of the Euclidean distances between all high-dimensional data points and the cluster center of their cluster, as the degree of difference between the high-dimensional data point and the high-dimensional data points in its cluster. This method not only calculates the Euclidean distance between a single high-dimensional data point and its cluster center, but also introduces the average Euclidean distance between all high-dimensional data points and the cluster center of their cluster, combining local (the difference between a single data point and the cluster center) and global (the average difference between all data points) information, making the assessment of the degree of difference more comprehensive. The introduction of the mean provides a relative benchmark for the difference of each data point, helping to identify abnormal points with large deviations within the cluster.

[0007] Furthermore, the degree of difference is calculated, including: the degree of difference is also negatively correlated with the standard deviation of the Euclidean distances between all high-dimensional data points and the cluster center in the cluster where the high-dimensional data point is located. The standard deviation reflects the degree of dispersion of the Euclidean distances between the data points and the cluster center within the cluster. When the standard deviation is large, it means that the data distribution within the cluster is relatively dispersed. At this time, the large Euclidean distance between a single data point and the cluster center may not be abnormal, but part of the normal fluctuation within the cluster. By making the degree of difference negatively correlated with the standard deviation, the misjudgment of normal data points in this case can be effectively reduced, thereby improving the robustness and noise resistance of the difference assessment.

[0008] Furthermore, the load variation degree of high-dimensional data points is calculated using the following formula: Where, Indicates the The first The load change weight of high-dimensional data points; Indicates the The total number of high-dimensional data points in the clusters; 、 represents the ordinal number of high-dimensional data points, ; Indicates the In the clusters, high-dimensional data points and The collection time interval of high-dimensional data points, Represents the standard normalization function.

[0009] Furthermore, the optimal K value of each high-dimensional data point is calculated using the following formula: Where, Indicates the The first The optimal high-dimensional data point value, For preset value, Indicates the The first The degree of load change of high-dimensional data points, The optimal K value is directly linked to the load variation, allowing for dynamic adjustment based on the specific operating status of each data point.

[0010] Furthermore, all high-dimensional data points are clustered, including using an iterative self-organizing clustering algorithm to cluster all high-dimensional data points and obtain clustering results. Iterative self-organizing clustering algorithms effectively handle the complexity of high-dimensional data by mapping high-dimensional data into a low-dimensional space (such as a two-dimensional grid) or by iteratively optimizing cluster centers. This dimensionality reduction or clustering process preserves the key characteristics of the data while reducing computational complexity, making subsequent calculations of variance and anomaly scores more efficient and accurate.

[0011] Furthermore, the optimized LOF algorithm is used to calculate the anomaly score of each high-dimensional data point, including: for any high-dimensional data point, using its corresponding optimal K value, and using the optimized LOF algorithm to calculate the anomaly score of the high-dimensional data point.

[0012] Furthermore, the operation status is monitored using the abnormality score, including: if the abnormality score of the high-dimensional data point is greater than an abnormality threshold, an alarm is issued, wherein the empirical value of the abnormality threshold is 0.7.

[0013] In a second aspect, the present invention further provides a modular intelligent box-type substation operation monitoring system, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above modular intelligent box-type substation operation monitoring methods.

[0014] Beneficial effects:

[0015] (1) The K value of the LOF algorithm is optimized and dynamically adjusted through cluster analysis and load variation calculation. This solves the problem of inaccurate detection when the load fluctuates due to the fixed K value in the traditional LOF algorithm, improves the sensitivity and accuracy of anomaly detection, and avoids misjudgments caused by changes in the data fluctuation range.

[0016] (2) High-dimensional feature points are constructed by integrating multi-dimensional data such as voltage, current, and temperature, and a time interval weighting factor is introduced to capture the temporal characteristics of sudden load changes. By jointly calculating intra-cluster differences and temporal dispersion, steady-state and transient anomalies are effectively distinguished, enhancing the ability to identify local sudden anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. 4 is a flow chart schematically illustrating a method for calculating an optimal K value according to an embodiment of the present invention.

[0018] Figure 2 FIG. 4 is a flow chart schematically illustrating a method for calculating a load variation degree of a high-dimensional data point according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] In one embodiment, a box-type substation is a compact, integrated substation device, typically used in urban or rural power distribution systems. It integrates transformers, switchgear, protection devices, and distribution equipment into a closed metal box. Box-type substations are small, compact, and easy to install and maintain. They are suitable for locations with low loads or requiring rapid deployment, such as residential communities, industrial parks, commercial buildings, and transportation hubs. Monitoring the operation of box-type substations can improve substation reliability, extend equipment life, and ensure the stability and security of power supply.

[0021] like Figure 1 As shown, S101: collecting operation data of the substation.

[0022] In one embodiment, the operating data of the substation is collected at a preset collection frequency, where the operating data includes load current, load voltage, and temperature data. Specifically, current sensors, voltage sensors, and temperature sensors can be used to collect the load current, load voltage, and temperature data during the operation of the substation, respectively, and the data of different dimensions collected at the same time are used as a high-dimensional data point.

[0023] S102: Calculate the load variation degree of each high-dimensional data point.

[0024] In one embodiment, during the operation of the substation, due to the different electricity consumption behaviors of users, the load changes in different time periods are different, and the load will change during its continuous operation. When the load is high, the substation has a higher demand for each dimension, and the data of each dimension fluctuates greatly; while when the load is low, the substation has a lower demand for each dimension, and the data of each dimension fluctuates less. Therefore, during the operation of the substation, its load will continue to change. The constantly changing load will cause the results of abnormality detection using the fixed K value of the traditional LOF algorithm to be inaccurate. Therefore, the present invention needs to analyze the load changes of the substation to facilitate the subsequent calculation of different K values ​​based on the different load change degrees of each high-dimensional data point.

[0025] It should be noted that since substations experience both high and low loads during operation, a clustering algorithm can be used to cluster the collected substation load current, load voltage, and temperature data to facilitate analysis. This facilitates subsequent analysis based on the clustering results. Specifically, an iterative self-organizing clustering algorithm is used to cluster the collected high-dimensional data points to obtain clustering results. The clustering results include multiple clusters and the high-dimensional data points within each cluster.

[0026] In one embodiment, the load change degree of each high-dimensional data point is calculated, including: taking the product of the difference degree of any high-dimensional data point and the load change weight of the high-dimensional data point as the load change degree of the high-dimensional data point, wherein the load change weight of the high-dimensional data point is positively correlated with the collection time interval between the high-dimensional data point and all high-dimensional data points in the cluster to which it is located, and the difference degree reflects the distance between the high-dimensional data point and the high-dimensional data points in the cluster to which it is located.

[0027] like Figure 2 As shown, S1021: calculate the degree of difference.

[0028] In one embodiment, for any high-dimensional data point in any cluster, if the distance between the high-dimensional data point and the high-dimensional data points within the cluster is large, it indicates that the high-dimensional data point has a large intra-cluster difference. During clustering, similar high-dimensional data points need to be assigned to the same cluster. Therefore, if the intra-cluster difference of the data point is large, it indicates that the load variation of the high-dimensional data point is large. Based on the above characteristics, the degree of difference between any high-dimensional data point and the high-dimensional data points in its cluster is calculated.

[0029] In one embodiment, the degree of difference is calculated by calculating the Euclidean distance between any high-dimensional data point and the cluster center of the cluster in which it is located, and the average of the Euclidean distances between all high-dimensional data points and the cluster center of the cluster in which they are located, as the degree of difference between the high-dimensional data point and the high-dimensional data points in the cluster in which it is located. The calculation formula for the degree of difference is: Where, Indicates the The first The degree of difference between a high-dimensional data point and its cluster; Indicates the The first The Euclidean distance between a high-dimensional data point and the cluster center of the cluster it belongs to; Indicates the The total number of high-dimensional data points in the clusters; Represents the ordinal number of high-dimensional data points. It represents the ratio of the Euclidean distance between the high-dimensional data point and the cluster center of the cluster to the mean of the Euclidean distances between all high-dimensional data points and the cluster centers of the clusters they belong to. The larger the value, the greater the intra-cluster difference of the high-dimensional data point.

[0030] In another embodiment, another method for calculating the degree of difference is provided, which uses the standard deviation of the Euclidean distance between all high-dimensional data points in the cluster and the cluster center to weight the above formula. The calculation formula is: Where, Indicates the The first Intra-cluster differences of high-dimensional data points; Indicates the The first The Euclidean distance between a high-dimensional data point and the cluster center of the cluster it belongs to; Indicates the The total number of high-dimensional data points in the clusters; Represents the ordinal number of high-dimensional data points; Indicates the The standard deviation of the Euclidean distance between all high-dimensional data points in a cluster and the cluster center; Expressed as a natural constant The reason is that: It represents the standard deviation of the Euclidean distance between all high-dimensional data points and the cluster center in the cluster. The smaller the value, the more consistent the distribution of high-dimensional data points in the cluster, and the higher the confidence level of the intra-cluster difference of the high-dimensional data points. It avoids the problem of inaccurate calculation results due to a large ratio (the ratio refers to the ratio of the Euclidean distance between a high-dimensional data point and the cluster center of its cluster to the mean of the Euclidean distances between all high-dimensional data points and the cluster center of their cluster), but the distribution of high-dimensional data points in the cluster is relatively discrete.

[0031] S1022: Calculate the load change weight of the high-dimensional data point.

[0032] In one embodiment, during substation operation, load changes are often sudden changes caused by different users' electricity usage behaviors. Therefore, local changes in high-dimensional data points need to be considered. However, the above clustering method, which clusters current, voltage, and temperature data together, does not consider time factors. Therefore, this embodiment calculates the load change weight by calculating the time interval of high-dimensional data points in any cluster, thereby more accurately calculating the load change degree of high-dimensional data points.

[0033] In one embodiment, for any high-dimensional data point, the high-dimensional data point is used as the target data point, the collection time of all high-dimensional data points in the cluster where the high-dimensional data point is located is obtained, and the collection time interval between the target data point and all data points in the cluster is calculated (the absolute value of the difference in collection time); and the load change weight is calculated based on the collection time interval, and the calculation formula is as follows: Where, Indicates the The first The load change weight of high-dimensional data points; Indicates the The total number of high-dimensional data points in the clusters; 、 represents the ordinal number of high-dimensional data points, ; Indicates the In the clusters, high-dimensional data points and The collection time interval of high-dimensional data points; Represents the standard normalization function. The longer the time interval between the target data point and the high-dimensional data points in its cluster, the greater the load change weight of the target data point, indicating that there are no high-dimensional data points with similar values ​​near the target data point's collection time.

[0034] S103: Calculate the optimal K value for each high-dimensional data point.

[0035] It should be noted that for any high-dimensional data point, the greater the load change of the high-dimensional data point, the wider the fluctuation range of the voltage, current, temperature and other parameters of the substation at that moment, the more dispersed the data point is, and there are high-dimensional data points with significant differences in the local neighborhood. In order to avoid the local density distortion caused by the small K value in this case, and mistakenly judge normal fluctuations as abnormalities, the greater the load change, the larger the K value.

[0036] In one embodiment, the optimal K value for each high-dimensional data point is calculated using the following formula: Where, Indicates the The first The optimal K value for high-dimensional data points, is the preset K value, Indicates the The first The degree of load change of high-dimensional data points, represents the ceiling function. The K value is preset based on experience, and the implementer can adjust it according to the specific implementation situation, for example .

[0037] S104: Detecting the operating status of the substation using the optimized LOF algorithm.

[0038] In one embodiment, for any high-dimensional data point, the optimal K value corresponding to it is used to calculate the anomaly score of the high-dimensional data point using the LOF algorithm. If the anomaly score of the high-dimensional data point is greater than the anomaly threshold, it indicates that the substation's operating status has been abnormal, thereby triggering an alarm and providing a warning to the staff. If the anomaly score of the high-dimensional data point is less than or equal to the anomaly threshold, it indicates that the substation's operating status is normal and operations can continue. The empirical value of the anomaly threshold is 1.4. In other embodiments, the empirical value of the anomaly threshold can also be 1.3 or 1.25, etc., and the implementer can adjust it according to the specific implementation situation.

[0039] The present invention also provides a modular intelligent box-type substation operation monitoring system. The system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, the modular intelligent box-type substation operation monitoring method according to the first aspect of the present invention is implemented.

[0040] The system further includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and thus will not be described in detail here.

[0041] In the present invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented using computer-readable / executable instructions stored or otherwise retained on such a computer-readable medium.

[0042] The above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.

Claims

1. A modular intelligent box-type substation operation monitoring method, characterized in that: include: Collect substation operation data and treat multiple operation data collected at the same time as a high-dimensional data point; The optimized LOF algorithm is used to calculate the anomaly score of each high-dimensional data point, and the operating status is monitored based on the anomaly score. The optimized LOF algorithm includes: calculating the optimal K value of each high-dimensional data point according to the load change degree of each high-dimensional data point, satisfying: ; Indicates the The first The optimal K value for high-dimensional data points, is the preset K value, Indicates the The first The degree of load change of high-dimensional data points, represents the ceiling function; Calculating the load variation degree of each high-dimensional data point includes: clustering all high-dimensional data points to obtain multiple clusters, and calculating the difference between any high-dimensional data point and the high-dimensional data points in the cluster where it is located; The product of the difference degree of any high-dimensional data point and the load change weight of the high-dimensional data point is used as the load change degree of the high-dimensional data point, wherein the load change weight of the high-dimensional data point is positively correlated with the collection time interval between the high-dimensional data point and all high-dimensional data points in the cluster to which it belongs; the load change degree satisfies: ； Indicates the The first The load change weight of high-dimensional data points; Indicates the The total number of high-dimensional data points in the clusters; 、 represents the ordinal number of high-dimensional data points, ; Indicates the In the clusters, high-dimensional data points and The collection time interval of high-dimensional data points, Represents the standard normalization function.

2. The modular intelligent box-type substation operation monitoring method according to claim 1 is characterized in that: Calculating the degree of difference, including: Calculate the Euclidean distance between any high-dimensional data point and the cluster center of its cluster, and the average of the Euclidean distances between all high-dimensional data points and the cluster center of their cluster, as the degree of difference between the high-dimensional data point and the high-dimensional data points in its cluster.

3. The modular intelligent box-type substation operation monitoring method according to claim 1 is characterized in that: Calculating the degree of difference, including: The degree of difference is also negatively correlated with the standard deviation of the Euclidean distances between all high-dimensional data points and the cluster center in the cluster where the high-dimensional data point is located.

4. The modular intelligent box-type substation operation monitoring method according to claim 1 is characterized in that: Cluster all high-dimensional data points, including: All high-dimensional data points are clustered using an iterative self-organizing clustering algorithm to obtain clustering results.

5. The modular intelligent box-type substation operation monitoring method according to claim 1 is characterized in that: The optimized LOF algorithm is used to calculate the anomaly score of each high-dimensional data point, including: For any high-dimensional data point, the corresponding optimal K value is used to calculate the anomaly score of the high-dimensional data point using the optimized LOF algorithm.

6. The modular intelligent box-type substation operation monitoring method according to claim 1 is characterized in that: Monitoring the operating status using the abnormality score includes: If the anomaly score of the high-dimensional data point is greater than the anomaly threshold, an alarm is issued.

7. The modular intelligent box-type substation operation monitoring method according to claim 6 is characterized in that: The empirical value of the abnormal threshold is 0.

7.

8. A modular intelligent box-type substation operation monitoring system, comprising a processor and a memory, characterized in that: The memory stores a computer program, and the processor executes the computer program to implement the modular intelligent box-type substation operation monitoring method according to any one of claims 1 to 7.

Images