KR-102963350-B1 - The Method, System, And Computer-Readable Recording Medium That Synthesizes AMI Data Based on Segmented-STL

KR102963350B1KR 102963350 B1KR102963350 B1KR 102963350B1KR-102963350-B1

Abstract

The present invention relates to a segmented STL-based AMI data synthesis method, system, and computer-readable recording medium, wherein a plurality of AMI data within an AMI data group are clustered based on data patterns and data sizes; each of the plurality of AMI data is input into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; a distribution is generated for each of the trend components and residual components for each cluster group; noise derived from the distribution is applied to each of the trend components and residual components to modify them; and the existing seasonal components, modified trend components, and modified residual components are synthesized to generate synthesized AMI data.

Inventors

손성용
김민수

Assignees

가천대학교 산학협력단

Dates

Publication Date: 20260511
Application Date: 20250304
Priority Date: 20241115

Claims (15)

A segmented STL (Seasonal-Trend decomposition using Loess) based AMI data synthesis method performed on a computing system comprising one or more processors and one or more memories, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above AMI data synthesis method is, A cluster group division step of applying a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A detailed cluster division step for dividing the cluster group into detailed clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation step that inputs each of multiple AMI data into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; A distribution generation step of generating a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generating a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation step for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis step for generating multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above detailed cluster division step is, A data integration step for calculating a total amount of multiple AMI data by integrating each of the multiple time series AMI data belonging to each cluster group; and AMI data synthesis method comprising: a quartile division step of sorting AMI data belonging to each cluster group in order of the size of the total amount of AMI data to derive quartiles, and dividing each cluster group into four sub-clusters based on the quartiles.
delete
delete
delete
A segmented STL (Seasonal-Trend decomposition using Loess) based AMI data synthesis method performed on a computing system comprising one or more processors and one or more memories, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above AMI data synthesis method is, A cluster group division step of applying a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A detailed cluster division step for dividing the cluster group into detailed clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation step that inputs each of multiple AMI data into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; A distribution generation step of generating a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generating a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation step for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis step for generating multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above distribution generation step is, AMI data synthesis method comprising: a second distribution generation step of collecting residual components of AMI data belonging to each sub-cluster and inputting them into a Generalized Extreme Value (GEV) module that provides predictions for extreme values of the data to calculate a second distribution for each sub-cluster.
A segmented STL (Seasonal-Trend decomposition using Loess) based AMI data synthesis method performed on a computing system comprising one or more processors and one or more memories, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above AMI data synthesis method is, A cluster group division step of applying a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A detailed cluster division step for dividing the cluster group into detailed clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation step that inputs each of multiple AMI data into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; A distribution generation step of generating a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generating a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation step for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis step for generating multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above component transformation step includes a trend component transformation step, and The above trend component deformation step is, A trend component noise derivation step of deriving a plurality of noises for a corresponding trend component from values sampled within a preset confidence interval from a first distribution of a sub-cluster to which each trend component belongs; and AMI data synthesis method comprising: a step of deriving a modified trend component, wherein a modified trend component according to a plurality of time series is derived by applying one noise randomly extracted from a plurality of noises derived from a first distribution of a sub-cluster to which each trend component belongs for each time unit to each of the trend components according to the plurality of time series.
delete
A computing system comprising one or more processors and one or more memories, and performing a segmented STL (Seasonal-Trend decomposition using Loess)-based AMI data synthesis method, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above computing system is, A cluster group division unit that applies a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A sub-cluster division unit that divides the cluster group into sub-clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation unit that inputs each of multiple AMI data into an STL separation module and separates them into seasonal components, trend components, and residual components according to time series; A distribution generation unit that generates a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generates a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation unit for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis unit that generates multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above detailed cluster division is, A data integration unit that calculates a total amount of multiple AMI data by integrating each of the multiple time series AMI data belonging to each cluster group; and A computing system comprising: a quartile partitioning unit that sorts AMI data belonging to each cluster group in order of the size of the total amount of AMI data to derive quartiles, and divides each cluster group into four sub-clusters based on the quartiles.
delete
delete
delete
A computing system comprising one or more processors and one or more memories, and performing a segmented STL (Seasonal-Trend decomposition using Loess)-based AMI data synthesis method, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above computing system is, A cluster group division unit that applies a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A sub-cluster division unit that divides the cluster group into sub-clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation unit that inputs each of multiple AMI data into an STL separation module and separates them into seasonal components, trend components, and residual components according to time series; A distribution generation unit that generates a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generates a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation unit for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis unit that generates multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above distribution generation unit is, A computing system comprising: a second distribution generation unit that collects residual components of AMI data belonging to each sub-cluster, inputs them into a Generalized Extreme Value (GEV) module that provides predictions for extreme values of the data, and calculates a second distribution for each sub-cluster.
A computing system comprising one or more processors and one or more memories, and performing a segmented STL (Seasonal-Trend decomposition using Loess)-based AMI data synthesis method, An AMI data group, comprising multiple AMI data collected according to a time series by an AMI, is stored in the above computing system, and The above computing system is, A cluster group division unit that applies a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A sub-cluster division unit that divides the cluster group into sub-clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation unit that inputs each of multiple AMI data into an STL separation module and separates them into seasonal components, trend components, and residual components according to time series; A distribution generation unit that generates a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generates a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation unit for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis unit that generates multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above component deformation part includes a trend component deformation part, and The above trend component deformation part is, A trend component noise derivation unit that derives multiple noises for a corresponding trend component from values sampled within a preset confidence interval from a first distribution of a sub-cluster to which each trend component belongs; and A computing system comprising: a derivation of
delete
A computer-readable recording medium for implementing a segmented STL (Seasonal-Trend decomposition using Loess)-based AMI data synthesis method performed on a computing system comprising one or more processors and one or more memories, The above computing system stores an AMI data group in which multiple AMI data measured according to a time series by an AMI are collected, and The above computer-readable recording medium comprises computer-executable instructions that cause the computing system to perform the following steps, and The steps below are: A cluster group division step of applying a plurality of AMI data within an AMI data group to a preset clustering algorithm to divide the AMI data group into a plurality of cluster groups; A detailed cluster division step for dividing the cluster group into detailed clusters based on the size of AMI data according to the time series belonging to each cluster group; An AMI data component separation step that inputs each of multiple AMI data into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; A distribution generation step of generating a first distribution for each sub-cluster by collecting trend components of AMI data belonging to each sub-cluster, and generating a second distribution for each sub-cluster by collecting residual components of AMI data belonging to each sub-cluster; A component transformation step for multiple AMI data sets, deriving a modified trend component according to a time series by applying noise according to a first distribution of the sub-cluster to which the AMI data belongs for each preset time unit to a trend component according to a time series, and deriving a modified residual component according to a time series by applying noise according to a second distribution of the sub-cluster to which the AMI data belongs for each time unit to a residual component according to a time series; and A data synthesis step for generating multiple synthesized AMI data by synthesizing a seasonal component according to the time series of each of the multiple AMI data, a deformation trend component according to the time series, and a deformation residual component according to the time series; The above detailed cluster division step is, A data integration step for calculating a total amount of multiple AMI data by integrating each of the multiple time series AMI data belonging to each cluster group; and A computer-readable recording medium comprising: a quartile division step of sorting AMI data belonging to each cluster group in order of the size of the total amount of AMI data to derive quartiles, and dividing each cluster group into four sub-clusters based on the quartiles.

Description

The Method, System, and Computer-Readable Recording Medium That Synthesizes AMI Data Based on Segmented-STL The present invention relates to a segmented STL-based AMI data synthesis method, system, and computer-readable recording medium, wherein a plurality of AMI data within an AMI data group are clustered based on data patterns and data sizes; each of the plurality of AMI data is input into an STL separation module to separate them into seasonal components, trend components, and residual components according to time series; a distribution is generated for each of the trend components and residual components for each cluster group; noise derived from the distribution is applied to each of the trend components and residual components to modify them; and the existing seasonal components, modified trend components, and modified residual components are synthesized to generate synthesized AMI data. Advanced Metering Infrastructure (AMI) is an intelligent metering system that collects and manages energy data, such as electricity, heat, and gas, in real time, and is one of the core technologies for building smart grids. The AMI data collected serves as a valuable resource that extends beyond simple energy usage information to enable various analyses and predictions. To effectively utilize the massive volume of AMI data generated, appropriate processing and analysis methods that reflect the characteristics of the data are required. Meanwhile, AMI data contains individual users' energy consumption patterns, which raise the possibility of inferring sensitive information such as lifestyle patterns, residence status, and activities during specific time periods. If this data is utilized without sufficient protective measures, there is a risk of infringing upon personal privacy, and security threats may arise in the event of a data leak. Therefore, to safely utilize AMI data, technology capable of guaranteeing personal information protection during the data analysis and storage processes is required, highlighting the need to strike a balance between data utilization and security. Conventional methods for protecting AMI data utilize data anonymization and encryption technologies. Anonymization involves removing or modifying elements from data that can identify individuals to prevent the direct exposure of personal information, while encryption applies cryptographic algorithms to maintain security during data transmission and storage. However, conventional methods cannot completely eliminate the time-series characteristics and patterns inherent in AMI data, leaving a possibility for re-identification of individuals; furthermore, encrypted data may face difficulties in sharing and utilizing for research. Consequently, there is a need to develop data processing technologies that protect personal information that may be exposed from the data while enabling its effective use for analysis and research. FIG. 1 schematically illustrates the components of a computing system that performs a segmented STL-based AMI data synthesis method according to one embodiment of the present invention. FIG. 2 schematically illustrates the steps of a segmented STL-based AMI data synthesis method according to one embodiment of the present invention. FIG. 3 schematically illustrates the process of dividing an AMI data group into a plurality of cluster groups according to one embodiment of the present invention. FIG. 4 schematically illustrates the process of calculating the total amount of AMI data and sorting the AMI data in order of the size of the total amount of AMI data according to an embodiment of the present invention. FIG. 5 schematically illustrates the process of dividing each of a plurality of cluster groups into sub-clusters according to one embodiment of the present invention. FIG. 6 illustrates AMI data with components separated by an STL separation module according to an embodiment of the present invention. FIG. 7 schematically illustrates the process of calculating a first distribution for each detailed cluster by a GPR module according to one embodiment of the present invention. FIG. 8 schematically illustrates the process of calculating a second distribution for each detailed cluster by a GEV module according to one embodiment of the present invention. FIG. 9 schematically illustrates the process of performing a component modification step according to one embodiment of the present invention. FIG. 10 illustrates an exemplary process of generating synthetic AMI data by synthesizing a seasonal component, a deformation trend component, and a deformation residual component according to one embodiment of the present invention. FIG. 11 illustrates the internal configuration of a computing device according to one embodiment of the present invention. Hereinafter, various embodiments and/or aspects are disclosed with reference to the drawings. For illustrative purposes, numerous specific details are disclosed in the following description