Event-driven forecasting of wholesale electricity price and frequency regulation price using machine learning algorithms

Abstract

The wholesale electricity market is composed of real-time market and procurement. Since the fully liberalization of the energy market in Singapore in 2018, competition among the market participants has become intensive. Therefore, the forecast of the electricity price, including both Uniform Singapore Energy Price (USEP) and regulation price (REG), is essential for market participants to trade at the best possible price. This study proposes an event-driven forecast model to provide a real-time automatical price forecast of the wholesale electricity price using the cutting-edge regression tree ensembled algorithm, XGBoost. The model is triggered every half of an hour by the dispatch of the latest electricity price, which will be also used as the input together with other necessary training data from the historical price database. The training is finished within 1–2 min and the model makes the predictions on prices of the next 6 h. A better prediction has been demonstrated benchmarked to the forecast provided commercially by Energy Market Company. The forecast model only takes commercially available data as the input and is embedded with messaging broker, enabling online collaboration of different modules. The model is readily implemented, and thus can serve as an effective supplementary tool in virtual power plant operation with a higher renewable energy penetration in the future.