With the increasing share of renewable energy in the electric sector, new power plants are forced to provide grid services in order to ensure a power grid operation. Among these requirements, this project is focused on facilitating the management of the power reserves in photovoltaic power plants taking into account its variability due to climate conditions such as clouds. Managing these power reserves gives flexibility to these power plants and therefore also to the electric grid.
So, the main goal of this Master Thesis is to develop a model based on Artificial Intelligence technology that is capable of estimating real-time maximum power of a large-scale photovoltaic power plant (LS-PVPP). Knowing the maximum power that a plant is able to produce helps to be aware of the power reserves when it is working in curtailment mode and manage them according to TSO requirements.
To build up the model, real data of a photovoltaic power plant is used after experimenting a pre-treatment step. The methodology to estimate the available power is inspired by a high accurate technique developed by NREL which consists of using several inverters as power reference in order to estimate the total power of the photovoltaic plant. The model developed in this project uses Artificial Neural Networks to estimate the maximum power, whereas NREL method uses linear relationships. Both techniques are compared so as to determine the accuracy of the neural network’s estimated values. The neural network model is able to estimate with higher performance than NREL’s method. This leads to affirm that neural networks can be used to estimate the maximum power of large-scale photovoltaic plants.
Furthermore, a brief study of the influence of the size of the power plant on the performance of the neural network model is done. According to the results, smaller photovoltaic power plants have more difficulties to implement neural networks to estimate its available power than L-S PVPP.
