DYNA JOURNAL ENGINEERING

Key words:

wind power forecasting, multivariate time series, deep learning, convolutional neural networks, attention mechanism, transformer models, renewable energy prediction, SCADA data, hybrid neural networks, error reduction, day-ahead forecasting, energy systems modelling

Article type:

ARTICULO DE INVESTIGACION / RESEARCH ARTICLE

Section:

RESEARCH ARTICLES

Researchers have put forward that the variations in the climate based on various weather conditions directly affect wind power forecasting. Predicting weather changes and wind power output accurately and theoretically using statistical prediction models is complex. With conventional learning models, forecasting long-term wind power can be made to work with mean absolute percentage error of ten per cent to seventeen per cent; this did not meet our renewable energy project's engineering requirements. The Generative Convolutional Information Encoding Network model (GCIENM) is proposed to achieve the correlations among power generation and meteorological parameters. In the wind power forecasting field, the presented technique has broad applicability. Henceforth, the research study focuses on the long-term, one day to three days ahead in wind power prediction with the (MAPE) of below 10% by employing GCIENM-based ATM and MVPNN. When we experimented, the GCIENM model performed better using the outputs of wind power generation in a wind power plant located in Scada and other historical weather data. The production of the experiment shows a MAPE value of 6% in the prediction of wind power in three days; for our project, this Value is sufficient for our requirement. This work finally compared the performance based on the proposed prediction model's performance for power forecasting. Our experiment showed that the GCIENM performs better than the three other models for predicting wind power by the parameters of forecast accuracy, error reduction stability and data input volume.
Keywords- renewable energy, prediction, deep learning, error rate, actual prediction