In this work, it was investigated how a production requirement-based model and a system-identified model, both using first-order transfer functions, can be combined. The study developed a tool using Python programming to simulate the expected delivery of Frequency Containment Reserve based on production requirements. Simultaneously, system identification is used to analyze the actual delivery, enabling assessment of a hydropower unit's capacity to provide FCR based on operational data. The focus is on identifying system constants and analyzing their impact on system performance in various operational scenarios. By applying frequency filtering to the dataset of actual delivered active power, the study aims to establish a fair comparison between expected and actual deliveries concerning production requirements. The study describes the construction of the model and the theoretical attributes supporting its design.
The results from the study present three operational scenarios in which the production requirement-based model simulates the expected active power delivery, while the system-identified model simulates the actual active power delivery from the ancillary services FCR-N, FCR-D upward, and FCR-D downward. The results indicate that the hydropower units' dynamic performance requirements can be assessed for FCR-N with this method but not for FCR-D. Static performance requirements can be evaluated for both reserves, albeit with varying levels of uncertainty for both FCR-N and FCR-D. The results also emphasize the importance of combining the production requirement-based model with the system-identified model to identify periods suitable for analysis.
The discussion highlights the results of the various operational scenarios for the production requirement-based model, the system-identified model, and the actual delivered active power, while also reviewing the choice of method and its application. It is also discussed that there are certain challenges in trying to capture the complex dynamics of a hydropower plant with a simple transfer function. The challenges lie in handling nonlinear effects and rapid regulations that can affect the model's accuracy. It has also been discussed that by expanding the models with additional production requirements and factors, FCR-D could be analyzed more effectively and with greater success.
The conclusions indicate that a combination of theoretical modeling, system identification, and signal processing is an efficient way to use Python to analyze the characteristics relevant to production requirements for FCR delivery using operational data. Furthermore, the models tested in this study show potential for further development. This suggests that the work presented here may be a good starting point for a concept that seems to work well and that, with additional improvements and refinements, could lead to even more robust and reliable methods for the analysis and optimization of FCR delivery.
2024.