Selected Tools For Wind Farm Output Daily-Hourly Forecasting For The Wholesale Market

Abstract

Owners of generating sources after their connection to the power grid become participants of the electricity market, including the balancing market. From that moment on, each participant is obliged to forecast their own generating units' output with a specified advance time. The adopted energy transformation policy removes investment restrictions, in particular concerning the so-called distance act (10H), which stopped the dynamic development of this technology on land. This approach will contribute to the construction of more wind farms. The greater the number of generation sources dependent on weather conditions, the more difficult their predictability and the greater the risk of trade imbalances in participants ' purchasing portfolios. Every incorrect energy forecast that differs from the actual output will result in higher costs of participation in the market. Effective output forecasting allows companies to reduce the cost of their participation. This paper presents a method of forecasting wind farm output using artificial neural networks, which can be an alternative tool for analytical and statistical models. This paper aims to evaluate the effectiveness of the farm output forecast model, i.e. the output modelling for specific weather conditions. The paper presents the farm output affecting factors that should be included in the model, and it shows that the neural network can reproduce farm output curves similar to the catalogue curves with consideration of the object's characteristics. The author has exhaustively researched the subject for 10 years, reaching the main conclusion that it is impossible to create one universal forecast model for every farm. This means that each such facility requires an individual approach to obtain an effective forecast. The changing market environment requires further action and the development of new models suitable for the needs of the markets within a shorter time horizon.