A new approach on conducting flexibility studies based on neural networks - application on the Hellenic transmission system

In order to meet the environmental targets, a large scale integration of renewable energy sources (RESs) has been taking place in power systems. Although those sources contribute to more clean energy and help to reduce the green house gasses emissions, they pose significant challenges to the system’s stability due to the intermittency and the high level of dependence to the weather conditions. One of the main challenges is to sustain the system’s balance and to guarantee the flexibility adequacy. For this purpose, the need for more comprehensive flexibility studies will increase in the future, since different scenarios should be examined according to various level of RESs, demand and weather conditions. In this work, we propose a novel approach for the unit commitment (UC) algorithm that is implemented in the conventional flexibility studies that aims to reduce the computational time and simultaneously sustain the results’ accuracy in a satisfactory level. For this scope, we designed a nested-neural network system of two recurrent neural networks with long short-term memory, which are integrated and aim to reproduce the results of the complex conventional UC algorithm. Results show that the predicted insufficient ramping resource expectation of the two solutions are similar, while the computational time is being reduced by almost 99%.