КОРОТКОСТРОКОВИЙ ІНТЕРВАЛЬНИЙ ПРОГНОЗ СУМАРНОГО ВІДПУСКУ ЕЛЕКТРОЕНЕРГІЇ ВИРОБНИКАМИ З ВІДНОВЛЮВАНИХ ДЖЕРЕЛ ЕНЕРГІЇ

• Date: 2019
• Main Authors: Блінов, І.В., Мірошник, В.О., Шиманюк, П.В.
• Format: Article
• Language: Ukrainian
• Published: Інститут електродинаміки Національної академії наук України 2019
• Subjects: відновлювані джерела енергії; ринок електричної енергії; короткострокове прогнозування; прогнозний інтервал; нейронні мережі глибинного навчання
• Online Access: https://prc.ied.org.ua/index.php/proceedings/article/view/115
• Journal Title: Proceedings of the Institute of Electrodynamics of the National Academy of Sciences of Ukraine

The paper proposes the architecture of deep learning artificial neural network for short-term forecasting of total electricity supply by renewable energy sources (RES) producers. The paper shown that using such neural network, it is advisable to predict 10 and 90 percentile of error distribution, which gives the lower and upper bounds of the forecast interval with a hit probability of 0.8 in addition to point forecasting of the most probable value of electricity release. An error function is used for training, which is a combination of the mean squared error deviation and the quantile regression error for the percentile model. The test of the model was carried out on the real data of the total release of RES producers published by SE “Energorynok”. The quality of the forecast was compared with forecast data of the manufacturers. The minimum average error is reached by combination of neural network and manufacturers' forecasts. The lowest maximum error is provided by the independent neural network forecast. The forecast interval provides a probability ​​of 0.82 for getting into this interval of actual values with an expected value of 0.8. References 9, figures 4, table.