| Abstract [eng] |
This project analyses the application of artificial intelligence in a hybrid power system, which consists of a (floating) photovoltaic power plant and a hydro pumped storage power plant. The hybrid power system is managed by forecasting the generation of a solar power plant and optimising the bidding strategy of a hybrid power system by maximising the revenue, taking into account the day ahead market and the imbalance prices, solar power plant and Baltic power system imbalance volumes. The analysis of one year hourly meteorological data was performed to select relevant features. Linear regression, nonlinear regression, artificial neural network, support vector machines, random forest and ensemble methods were applied to forecast the generation of the solar power plant. Due to the fact that the installed capacity of the energy storage is larger than the capacity of the generating unit in this hybrid power system, hourly operating schedule of the hydro pump storage power plant was optimised for each week in the year, and by taking it into the consideration, the optimal imbalance management strategy of each forecasting model was developed. The results indicate that the management strategy, developed based on the results of the empirical forecasting model, accounts for lower revenue than any of the artificial intelligence forecasting models analysed. It was also found that the management strategy, developed based on the results of the most accurate artificial intelligence forecasting model, does not necessarily provide the highest revenue. |
