Universidad Pontificia Comillas. Madrid (Spain)
March 16th, 2023
Summary:
Island power systems, and in general smaller power systems, have specific characteristics that make the transition to renewable energy generation even harder. There are different challenges that are hindering the operation of a bigger share of renewable energy sources. These challenges are acknowledged in this thesis and novel solutions are introduced, that can help operators to ease the transition and also contributes to the current state-of-the-art. To tackle the reserve scarcity, the economic and technical impacts of storing renewable energy by deloading are assessed. Deloading enables renewable sources to offer some reserve as their headroom. The results show that providing online reserve by deloading is always beneficial for small power islands, and beneficial for medium and big islands when renewable energy is abundant. As renewable generators are uncertain, storing energy for periods with generation surplus and releasing it when necessary has been recognized and studied a lot in the literature. Storage devices should be scheduled alongside the thermal and renewable generators, to be used efficiently. A formulation is presented in this thesis, that makes it possible to include liquid air energy storage in the unit commitment problem. Another issue that small power systems have been suffering from and it's getting worse by increasing the share of renewable energy sources is inertia scarcity. Any contingency can lead to a fast frequency decay when inertia is low. To prevent contingencies that will cause poor frequency response, it's been tried to include frequency dynamics in the scheduling process. It's challenging because frequency dynamics are highly non-linear and non-convex, making it very hard to add them to the unit commitment problem, which is usually solved as a mixed integer linear programming problem. In this thesis frequency constrained unit commitment is proposed with the help of machine learning, which keeps the size of the unit commitment similar to the conventional formulation. That also makes it suitable for more computationally demanding unit commitment formulations like robust and stochastic methods. This is important because systems with a high share of renewable generation are also uncertain. A robust frequency constrained unit commitment formulation is proposed in this thesis, which uses logistic regression to learn the frequency-related constraints. Then a machine learning process to learn the frequency nadir after outages as a constraint for the unit commitment problem is introduced and compared with an analytical state-of-the-art formulation. Results show that the proposed method based on machine learning is as effective as the analytical methods while having a considerably lower run-time.
Descriptors: Electrical technology and engineering, Electricity applications, Power technology, Power generation
Citation:
M. Rajabdorri (2023), Island system operation with high degree of renewable energy resources: proposing solutions for smaller power systems to ease the transition to clean energy generation. Universidad Pontificia Comillas. Madrid (Spain).
