Abstract
From Portuguese historical data, a Kohonen's Self-Organizing Map (SOM) is used to classify the days of the year according to their load curve profile. The SOM Kohonen's algorithm is an unsupervised neural network clustering technique often used for pattern classification tasks that preserves the topology of data. Through the visualization of the data by projecting it into a bidimensional grid it is possible to identify the 'natural' clusters. The main objective of the classification is to separate the load demand by clusters for forecasting purposes. The classification/clustering via SOM will be the first phase of a so called Artificial Neural Networks hybrid model, where two or more ANN models are combined in order to draw more accurate predictions. The historical data used is the Portuguese continental load demand from 2009 to 2017 registered in periods of 15 min, comprising a total of 96 load registries per day. This data will be classified in different clusters according to its load curve profile.
Original language | English |
---|---|
Title of host publication | 2021 IEEE Electrical Power and Energy Conference, EPEC 2021 |
Publisher | Institute of Electrical and Electronics Engineers (IEEE) |
Pages | 220-224 |
Number of pages | 5 |
ISBN (Electronic) | 9781665429283 |
DOIs | |
Publication status | Published - 2021 |
Event | 2021 IEEE Electrical Power and Energy Conference, EPEC 2021 - Virtual, Online, Canada Duration: 22 Oct 2021 → 31 Oct 2021 |
Conference
Conference | 2021 IEEE Electrical Power and Energy Conference, EPEC 2021 |
---|---|
Country/Territory | Canada |
City | Virtual, Online |
Period | 22/10/21 → 31/10/21 |
Keywords
- artificial neural networks
- electric demand
- self-organizing Maps
- short-term load forecasting
- time series clustering