КОРОТКОСТРОКОВЕ ПРОГНОЗУВАННЯ ЦІН РИНКУ «НА ДОБУ НАПЕРЕД» УКРАЇНИ ДЛЯ ЗАВДАНЬ ЕКОНОМІЧНОЇ ДИСПЕТЧЕРИЗАЦІЇ МІКРОМЕРЕЖ

• Datum: 2025
• Hauptverfasser: Мірошник, В.О., Сичова, В.В.
• Format: Artikel
• Sprache: Ukrainisch
• Veröffentlicht: Інститут електродинаміки Національної академії наук України 2025
• Schlagworte: короткострокове прогнозування; ринок «на добу наперед»; ціна електроенергії; нейронні мережі; глибоке навчання; мікромережі
• Online Zugang: https://prc.ied.org.ua/index.php/proceedings/article/view/417
• Назва журналу: Proceedings of the Institute of Electrodynamics of the National Academy of Sciences of Ukraine

The study presents a comparative analysis of short-term day-ahead electricity price forecasting models based on deep learning artificial neural networks. The research is conducted under the conditions of a unified trading zone of the Integrated Power System of Ukraine, synchronized with ENTSO-E, where the intraday price dynamics are shaped by high demand volatility, changing generation structure, regulatory constraints, and the consequences of attacks on the energy infrastructure. These factors lead to the emergence of nonlinear and locally structured patterns in the price time series, which complicates the application of traditional statistical forecasting methods. The study evaluates five deep learning architectures — LSTM, GRU, Transformer, MLP, and CNN — with the aim of assessing their effectiveness in modeling the relative price normalized by the hourly regulatory cap. The results demonstrate that forecasting accuracy is strongly dependent on the model’s ability to capture local structural patterns and adapt to stochastic transitions between market operating regimes. The convolutional neural network provides the lowest forecast error, indicating its suitability for operational planning and economic dispatching of microgrids, where the accuracy of price forecasts directly determines the effectiveness of decision-making. Ref. 19, fig. 2, tables 2.