Estudio de la integración de fuentes no convencionales en el mercado eléctrico colombiano considerando variaciones meteorológicas estacionales

Abstract

The transition towards renewable energy sources is a current topic in the world that seeks to reduce environmental impacts and achieve more sustainable energy generation in the long term. In Colombia, the inclusion of renewable energies in the electricity market is an issue that has gained relevance in recent years, given that the country has great potential for clean energy generation, such as solar, wind and hydroelectric. In this context, this research focuses on evaluating the impact of the inclusion of renewable energies in the Colombian electricity market, considering possible changes in the availability of natural resources, analyzing aspects such as the stock price, level of reserves of the plants water and stock market response to critical meteorological events. During development, an analysis of the composition of the electricity market is carried out in terms of energy matrix, offer prices, forecasts and economic dispatch. In addition, the possible changes that may occur in market behavior and the possible risk conditions that may arise due to climate changes are evaluated. To carry out this research, historical data on electricity generation and demand are used, as well as analysis of possible scenarios and mathematical simulation models of both statistical types and machine learning or artificial intelligence. Models created by machine learning, such as neural networks, allow the association of various input variables with objective output variables, detecting patterns that could not be expressed through conventional mathematical models, being capable of predicting volatile signals such as offer prices of hydro generation units. Artificial intelligence models play a fundamental role in this research, especially for the forecast of electric energy demand in the short and medium term, as well as for obtaining credible offer prices for the proposed case studies. Machine learning methods for time series forecasting can detect periodic behavior in a signal, even when the time series is partially fragmented or incomplete, which is an advantage they have over statistical models which are also capable of identifying seasonality, but necessarily continuous signals. Through simulation in specialized optimization software, the ideal dispatch of each case study was run, automatically carrying out the respective water resource planning. The results of the simulations showed the shortcomings in the present market model for the short-term energy market in Colombia, since it was demonstrated that energy losses through inefficiencies in the use of water resources (dumping) , does not necessarily represent an economic loss for the bidders who use said resource, and that the generators have too much freedom to choose their offer price, if it is taken into account that hydroelectric technology makes up almost 66% of the entire energy matrix available in the country, and only through the large-scale inclusion of renewable energies in the market can greater competitiveness be achieved. It was also observed that there is still a great dependence on thermal generation technologies for critical meteorological scenarios such as the El Niño phenomenon, and that even with a great integration of renewable energies, it could be insufficient to satisfy the reliability and availability needs of the WITHOUT in the coming years. By fulfilling the objectives of this investigation and with the help of all the information collected throughout 2023, which was an inflexible year for the generation market in the country, it is prudent and necessary to conclude that it is advisable to carry out a reform regulatory framework for the Colombian stock market, which contemplates both current and future problems to respond forcefully to the environmental phenomena that have affected the country with a direct consequence on the price of energy.