Short-Term Wind Speed Forecasting for an Autonomous Hybrid Power Plant of a Traction Railway Substation

Abstract

The paper considers the problem of leveling the electrical load schedule of a traction railway substation using an autonomous hybrid power plant, which includes a wind power plant and a hydrogen energy storage device. Because of the high length of the railway sections between substations and the limited capacity of power transmission lines, when heavy trains move, there is a lack of electricity, which leads to a voltage drop and a decrease in train speed. The potential solution to the problem creating of an autonomous power plant connected between substations. It generates wind or solar energy and stores it in an energy storage system to use the stored energy to minimize voltage drops. Short-term forecasting of the generated power is necessary to control the energy storage system. The paper presents the results of wind speed forecasting for the project of creating an autonomous power plant for the section of the railway between Yaya and Izhmorskaya (Kemerovo region, Russia). We analyzed hourly data on wind speeds and directions for 15 years and built a neural network model to predict wind speed and direction for 1 and 6 hours ahead. The influence of the training set size and self-adaptation mode on the forecasting accuracy and the stability of the model on a horizon of several years have been studied.