Forecasting the production of electricity from renewable energy sources has become crucial, namely it enables network operators to more easily predict and balance the production and consumption of electricity. A reliable forecast is beneficial in the optimization of the dispatching of its controlled units, in the stability of the network, as well as in making decisions about the purchase and sale of electricity on the energy markets, thus contributing to greater profits.
The master's thesis deals with the possibilities of creating predictive models of solar power plant production based on weather data, with the help of a machine learning tree regression model and the PVLIB tool, which with its mathematical models serves as a good reference for simulating the operation of photovoltaic energy systems. The work at the end of the result compares the two and highlights the strengths and weaknesses of each of them. Namely, the goal is to achieve tangible results and accuracy comparable to the reference mathematical PVLIB model by creating a machine-learned model. In addition, the thesis examines the tools and methods used, the relevance of input weather data sources in the analysis, as well as optimization algorithms. Furthermore, through a practical case study, it demonstrates how machine learning can assist experts in manual forecasting, showcasing the benefits and contributions of leveraging machine learning techniques.
|
