Increasing real-time awareness in Smart Grids. Or, solving differential-algebraic equations faster
- Date: Tuesday 16 May 2017, 11:00 – 12:00
- Location: Mathematics Level 8, MALL 1, School of Mathematics
- Type: Leeds Applied Nonlinear Dynamics, Seminars, Applied Mathematics
- Cost: Free
Petros Aristidou, University of Leeds. Part of the Leeds applied nonlinear dynamics seminars series.
While for many decades electricity grids evolved slowly and with minor alterations, an evolutionary change has marked the last 15 years. Society is now pushing for a secure, continuous, supply of green and environmentally friendly electricity, at a low price, and - if possible - without investing in new infrastructure. These emerging electric power networks are labelled as "Smart Grids". However, the new developments are putting a huge strain on electric power systems, forcing them to operate under increased uncertainty, with lower security margins, and fewer resources; while serving a higher than ever energy demand. Under these circumstances, network operators are called upon to manage their systems based on information and analysis performed minutes, hours, or even days before. This leads to either operating sub-optimally with increased security margins or sacrificing reliability for a lower cost.
It is thus necessary to increase real-time awareness in power systems providing operators with always updated information, valid estimates on the current stability limits and secure operation regions. This can be achieved by developing computational tools with real-time or faster than real-time (look-ahead) simulation capabilities. These tools, formulate the mathematical model of power systems based on real-time measurements and then perform a series of high-performance dynamic simulations to extract the necessary information.
In this talk, we will first introduce the mathematical models describing power systems, consisting of large-scale, hybrid (mixed continuous-discrete), stiff, differential-algebraic equations, and discuss the challenges leading to decreased simulation performance. Then, we will bring together ideas from domain decomposition and parallel computing techniques to provide look-ahead simulation capabilities. In addition, some relaxation techniques are presented to further speed-up the parallel simulations. Examples using a large real-life system and a realistic system representative of the whole continental European synchronous grid will be shown.
We shall close by discussing continuing challenges and future research avenues.
Petros Aristidou, University of Leeds