Nio Nano- And Micro-Particles Prepared By Solvothermal Method – Good Catalysts for CO2 Methanation

<p>The rapid increase and volatility of electricity and gas prices on European markets over the period 2021-2022 has caused serious political and economic concerns in Europe and the global community. This has led to a renewed need for more accurate short-term energy price forecasting models and increased interest in the economic value of these forecasts. The increased price volatility in the day-ahead market complicates forecasting processes [1] due to methodological phenomena and variations in the generating capacity of renewable energy technologies, the volatility of consumer behaviour, and the predictability characteristics associated with seasonality. These factors influence the inherent volatility of the electricity price at current prices [2]. The main objective of the study is to develop an accurate and efficient methodology for electricity price forecasting in the Lithuanian electrical power system. The research of this work proposes a new forecasting model based on artificial neural networks (ANNs), which is able to mitigate the high volatility of historical prices and other input variables/data in the short term (from 24 h/day to 168 h/week). The model is based on a combination of multiple recurrent neural networks and multilayer perceptrons (the model consists of 7 separate neural networks). The new mathematical model combines artificial neural networks based on multiple recurrent neural networks and multilayer perceptrons with different architectures and characteristics, and with different activation functions. The model was implemented on the basis of an empirical study using realistic electricity and gas market prices, meteorological and consumer load data for the period 2020-2022. The annual day-ahead hourly (24-hour period) forecast errors equal to 0.192 of nRMSE and 0.1650 of nMAE [3] were determined. Similar hourly (24-hour time series) accuracy was observed at shorter time scales (weekly, monthly). This shows that the model/algorithm is robust to seasonal price fluctuations. This feature is very important when the model is intended to be used continuously, throughout the year, covering all seasons/seasons. Comparing our model with relevant benchmarks, we found that our model demonstrated better accuracy than the integrated long-term recurrent convolutional network (ILRCN) model with nMAE= 0.1981, although this benchmark [4] has outperformed a series of other approaches. The newly proposed electricity price forecasting model/algorithm demonstrated better accuracy than the integrated long term recurrent convolutional network (ILRCN) model with nMAE= 0.1981, although this benchmark has outperformed a series of other approaches. In conclusion, the proposed model could be promising in practice, helping independent power and reserve suppliers and electricity generators to participate competitively in the electricity and reserve markets and to better manage potential risks. It is expected that the proposed multi-neural network forecasting model will be further developed to improve the accuracy of price forecasts and to economically justify its usefulness for different market participants.</p>