Impact Evaluation of Feature Selection to Short-Term Load Forecasting Models considering Weather Inputs and Load History
| Autor: | L. N. Silva, L. F. Lopes, V. G. Negri, Alzenira da Rosa Abaide, M. Capeletti |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Artificial neural network
Operations research Mathematical model Computer science 020209 energy media_common.quotation_subject Autocorrelation Feature selection 02 engineering and technology Mutual information Variation (game tree) Term (time) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Quality (business) media_common |
| Zdroj: | 2019 54th International Universities Power Engineering Conference (UPEC). |
| DOI: | 10.1109/upec.2019.8893598 |
| Popis: | The definition of a Short-Term Load Forecasting (STLF) model is not trivial. In this horizon, the load is hardly impacted by some factors like consume seasonalities, different to workdays, weekends and holidays, for example, hour of the day and weather conditions, like temperature variation along the day. Moreover, the load dynamic between two regions could not be similar, because factors like economic activities. Then, although the homogeneous condition of mathematical models to SLTF, is fundamental a well define model inputs. It will impact directly on the quality of load forecasting results. Thus, the objective of this paper is to develop an analysis about feature selection, related to load history input and weather inputs, applied in a model of STLF. The ANN STLF model results will be compared with Autocorrelation and Mutual Information results, defining the optimum inputs arrange to base case. The base case is composed by two Brazilian Utilities, of different climate and economic zones, approaching these inputs dynamics to each case. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|