Date: December 3, 2012
Speaker: Patrick Panciatic, RTE
Location: 414 CEPSR
Hosted by: Prof. Javad Lavaei
The complexity of the Pan European transmission system is increasing. More and more generations based on renewable energy are installed in the system. Some are dispersed (PV in distribution system) or others far way from load centers (off shore wind). They are generally intermittent (day ahead forecasts are not very accurate). We have a lot of difficulties to build overhead power lines. People don't like their impact on the landscape and they are now afraid of hypothetical effect of EMF on health. We must use more complex solutions: numerous Phase Shifters Transformers, upgrades of existing corridors by using new conductors (ACSS), underground cables, HVDC links embedded in AC systems. The Single European Electricity market is an optimizer which maximizes the use of existing assets, pushing the system to its limits.
To operate such a very large and complex system, tools are needed to help operators to make decisions. These tools must provide optimal decisions to the operators. This optimization problem is quite challenging. The first challenge is the very large scale, the full European system must be taken into account, electrical phenomena don't stop at administrative borders, (10000 electrical buses, 2000 generators, 100 PSTs, 10 HVDC links, etc.). Due to the increasing level of uncertainty, decisions based only the most likely future state of the system becomes unrealistic at least for decision making in day ahead. More and more post-fault actions are implemented to control the system using flexible devices (PST, HVDC link, etc.). The optimal preventive actions must be computed taking into account these possible post-fault actions. A large amount of controllable components have discrete behaviors: startup/shutdown of generators, switch on/off of capacitor/reactor banks, switching of breakers, taps of PSTs,... This optimization problem is a very large probabilistic security constrained optimal power flow problem including discrete controls.
We want to find possible solutions to solve this problem using approximations and/or relaxations. The first classical approach is to use a linear approximation for the modeling of the system (DC approximation). The problem becomes a very large Mixed Integer Linear Programming (MILP) and efficient solvers are available. The probabilistic aspect is still challenging even using this linear approximation. We propose a step first towards risk based optimal decisions. This is a worst case approach to help the operators to screen the most severe contingencies. This worst case approach is a bi-level optimization equivalent to a MILP or sequence of MILPs. For non linear modeling, the full problem is intractable for large system (MINLP) and even not very well defined. The associated problems with fixed value for the discrete variables are not convex. We propose to use Mathematical Program with Equilibrium Constraints (MPEC) to find a local optimum. For very large system with a large amount of discrete variables, the problem is still very difficult to solve. We partially solve this problem without considering contingency by a sequence of smaller MINLPs for the modeling of the discrete behavior of components. This approach ensures the feasibility of the solutions.
We believe that new ideas about convex relaxations of this very challenging problem could be very useful.
Patrick Panciatici graduated from the French Ecole Supérieure d'Electricité (Supélec) in 1984. He joined EDF R&D in 1985 as a research engineer developing new methods for load forecasting. He then managed the EUROSTAG (time domain simulator) Project and CSVC (Coordinated Secondary Voltage Control) Project. He joined RTE, the French Transmission System Operator in 2003, where he participated in the creation of the "Methods and Support" internal R&D department. He was the head of a team that developed real-time and operational planning tools for RTE, and ensured operational support for the use of those tools from 2003 to 2011. He is currently a scientific advisor in the R&D-I department of RTE, a representative of RTE in PSERC, and is involved in several European Projects (PEGASE, OPTIMATE, TWENTIES, iTesla, e-HIGHWAY2050). He is also a member of CIGRE, IEEE, SEE, and the "R&D plan" ENTSO-E Working Group.