MODELING AND FORECASTING OF RENEWABLE ELECTRICITY PRODUCTION

Abstract

The object of the study is the processes of electricity generation by renewable energy sources (solar and wind power plants) under unstable meteorological factors. The article considers the results of solving the current problem of modeling and forecasting the generation of renewable energy sources (solar and wind). The relevance arises from the significant nonlinearity of RES generation, which creates risks of destabilization of power systems. Building accurate models is critically important for operational dispatching management and minimizing imbalances in power grids. The purpose of the work is to improve the quality of forecasting RES generation based on a comparative analysis of deductive approaches and inductive algorithms. The subject of the study is mathematical models, algorithms and software tools for forecasting time series of renewable energy generation processes in the power grid based on meteorological factors. In the course of the work, an intelligent system based on Streamlit with a data collection pipeline was created. The system is designed on the basis of a three-tier architecture, which provides a clear separation of the logic of data collection, mathematical processing and visualization. The system is implemented in the Python programming language with the involvement of a modern Data Science technology stack. Using the developed system, a study of three models was conducted on an independent sample using the MAE and RMSE metrics. A comparative analysis of the constructed models (artificial neural network, linguistic method and method of similar trajectories) was conducted on real meteorological and energy data. Based on the research results, conclusions were drawn regarding the conditions and feasibility of using certain models for modeling and forecasting renewable electricity production.