Objective of this master thesis was to design a model of local secondary voltage control with artificial neural networks and to verify its voltage control capability with simulations using a power system model with variable share of solar power plants.
Firstly, we described the basics of voltage control in power systems. Afterwards, methods for power systems analysis and test power system IEEE RTS were defined. Artificial neural networks and Levenberg-Marquardt learning algorithm, which was used for learning process, were also described. Training data was acquired with large number of simulations and then used for learning of artificial neural networks of local secondary voltage control, which were then integrated in power system model.
We observed local secondary voltage control for different seasons and different shares of solar power plants. Local secondary voltage control operated well, except when power system model was in a state not covered by training data.