Keywords
clustering; electricity consumption; load curves; patterns; time granularity
Start Date
27-6-2018 9:00 AM
End Date
27-6-2018 10:20 AM
Abstract
Nowadays, for electricity operators it is crucial to get a balance between generation & distribution of energy versus current demand, so a better characterization of load profiles would help operators to achieve that balance. Data mining techniques are frequently used to discover patterns of energy consumption in time-granularity. This consumption is affected for seasonal trends, weather, type of days and hourly blocks. This research is aimed to determine differentiated profiles by applying clustering techniques to divide data in groups labeled in relation with top-down levels in time-granularity: seasons (top level), typology of days and hourly blocks (down level). A database containing energy consumption and weather variables measured daily in hourly blocks from January 2004 until August 2017 in Northern Mexico was used during the experiments. Clustering results determined that energy consumption all year long can be characterized in five load profiles according to seasons and daily blocks: Summer/Working days (Mondays to Fridays), Summer/Saturdays, Summer/Sundays and Holidays, Rest of the year/Mondays to Saturdays and Rest of the year/Sundays and Holidays. A final grouping task in hourly granularity was developed within the five temporal profiles separately and post-processed by using the Traffic Light Panel (TLP) tool to help final-user to understand hourly demand thresholds according to the meaning of the colors of the TLP’s, respectively. Finally, results were validated with experts on the field and some works are now focused on the implementation of an API with short-term forecasting tools, by considering the load profiling discovered so far.
Characterization of electric energy consumption with clustering techniques: a case study in Northern Mexico
Nowadays, for electricity operators it is crucial to get a balance between generation & distribution of energy versus current demand, so a better characterization of load profiles would help operators to achieve that balance. Data mining techniques are frequently used to discover patterns of energy consumption in time-granularity. This consumption is affected for seasonal trends, weather, type of days and hourly blocks. This research is aimed to determine differentiated profiles by applying clustering techniques to divide data in groups labeled in relation with top-down levels in time-granularity: seasons (top level), typology of days and hourly blocks (down level). A database containing energy consumption and weather variables measured daily in hourly blocks from January 2004 until August 2017 in Northern Mexico was used during the experiments. Clustering results determined that energy consumption all year long can be characterized in five load profiles according to seasons and daily blocks: Summer/Working days (Mondays to Fridays), Summer/Saturdays, Summer/Sundays and Holidays, Rest of the year/Mondays to Saturdays and Rest of the year/Sundays and Holidays. A final grouping task in hourly granularity was developed within the five temporal profiles separately and post-processed by using the Traffic Light Panel (TLP) tool to help final-user to understand hourly demand thresholds according to the meaning of the colors of the TLP’s, respectively. Finally, results were validated with experts on the field and some works are now focused on the implementation of an API with short-term forecasting tools, by considering the load profiling discovered so far.
Stream and Session
Stream B: (Big) Data Solutions for Planning, Management, and Operation and Environmental Systems.
B3: Sixth Session on Data Mining as a Tool for Environmental Scientists (S-DMTES-2018)