Operational Forecasting of Wind Speed for an Self-Contained Power Assembly of a Traction Substation

Currently, the prospects of creating hybrid power assemblies using renewable energy sources, including wind energy, and storage systems based on hydrogen technologies are being considered. To control such an system, it is necessary to perform operational sources generation forecasting, particularly forecasting assemblies. Their production depends speed direction wind. The article presents results solving problem for a assembly project aimed at increasing capacity railway section between Yaya Izhmorskaya stations (Kemerovo region Russian Federation). Hourly data speeds directions 15 years have been analyzed, neural network model has built, compact architecture multilayer perceptron proposed short-term 1 6 hours ahead. that developed allows minimizing risks overfitting loss accuracy due changes in operating conditions over time. A specific feature this work stability investigation trained long-term observations changes, as well analysis possibilities improving regular further training newly available data. nature influence size sample self-adaptation its horizon several established. It shown order ensure high meteorological required.