Predicción de la demanda de energía eléctrica a corto plazo utilizando redes neuronales artificiales.
The implementation of artificial neuronal networks was carried out due to the problem that exists in the forecast of short-term electricity demand which affects the performance of load flow, safety analysis, hydrothermal coordination, preventive maintenance plan of generators and economic dispatch....
Wedi'i Gadw mewn:
| Prif Awdur: | |
|---|---|
| Awduron Eraill: | |
| Fformat: | bachelorThesis |
| Iaith: | spa |
| Cyhoeddwyd: |
2021
|
| Pynciau: | |
| Mynediad Ar-lein: | http://repositorio.utc.edu.ec/handle/27000/7903 |
| Tagiau: |
Ychwanegu Tag
Dim Tagiau, Byddwch y cyntaf i dagio'r cofnod hwn!
|
Ychwanegu Tag | Crynodeb: | The implementation of artificial neuronal networks was carried out due to the problem that exists in the forecast of short-term electricity demand which affects the performance of load flow, safety analysis, hydrothermal coordination, preventive maintenance plan of generators and economic dispatch. Therefore, a computational tool based on artificial neural networks was developed, through the MATLAB computer program, using real data from the PRI12 ‘Santa Rosa de Pichul ’-‘ San Gerardo ’feeder, which allowed to analyze the behavior of the hourly electric demand in order to obtain an optimal forecast. The program is established by the graphical interface (GUIDE) with three modules such as historical data, neural network training and electrical demand forecast, for this research study the ANN is composed by an input layer, a hidden layer (hyperbolic tangent activation function) and an output layer (linear activation function), using the Levenberg-Marquardt training algorithm by using 288 delays, 10 neurons, 80% training data, 10% validation data and 10% test data, thus obtaining adequate performance results of 1. 01 x 10 ^ -7, in the same way it presents the best adjustment to the behavior of the data series with a percentage error of 1.60%, finally on the basis of the statistical curves of the real and predicted values an error of 0.99% was obtained, finding a satisfactory short-term forecast. |
|---|