Forecasting of electricity demand is a fundamental requirement for the energy sector since from its results important decisions are taken. The areas involved are maintenance of electrical networks, demand growth, increased installed capacity, among others, whose lack of precision can take high economic costs. In this work, we propose a method based on backpropagation neural networks and election of key variables as inputs. The number of neurons in the hidden layer was optimized. To avoid the overtraining the best time range of data was defined. The results show that the method works particularly well for short-term forecasting (24 or 48 hours).
|Original language||English (US)|
|State||Published - 18 Jan 2018|
|Event||2017 IEEE International Autumn Meeting on Power, Electronics and Computing, ROPEC 2017 - Ixtapa, Mexico|
Duration: 8 Nov 2017 → 10 Nov 2017
|Conference||2017 IEEE International Autumn Meeting on Power, Electronics and Computing, ROPEC 2017|
|Abbreviated title||ROPEC 2017|
|Period||8/11/17 → 10/11/17|