Energy sustainability is a key challenge of the 21st century. To effectively address the negative impacts of climate change without making sacrifices in terms of our quality of life, a transition towards renewable energy sources (RES) must take place. However, this transition poses a new challenge to the transmission system operators (TSO) around the globe as most technologies used to extract energy from renewable sources (solar panels, wind turbines etc.) substantially decrease the stability of conventional power systems, increasing the risks of blackouts and raising the overall costs of electricity.
A key concept related to the transient stability of a power network is inertia: this is a feature which, like its mechanical counterpart, ensures that the system does not undergo sudden and radical changes in its equilibrium state.
To efficiently manage a power grid and prevent undesired events such as blackouts and interruptions of service, the operators must constantly estimate the level of inertia of a power system, which is not directly measurable.
In our project we developed a new estimation method which tries to combine the best of both worlds: using a brand new convolutional neural network (CNN) architecture of our own making, named “StepNet”, which takes as input variables that can easily be measured by most TSOs, and outputs the estimate of global inertia of the power grid.