Centrum Wiskunde & Informatica
Founded in 1946, the Centrum Wiskunde & Informatica (CWI) is the national research institute for mathematics and computer science in the Netherlands. It is located at the Science Park Amsterdam and is part of the “Netherlands Organisation for Scientific Research” (NWO). The institute is internationally focused and renowned for its high quality research. Over 160 researchers conduct pioneering research and share their acquired knowledge with society. More than 30 researchers are employed as professors at universities. The institute has generated twenty-one spin-off companies.

CWI maintains excellent relations with industry and the academic world, both in the Netherlands as well as abroad. After their research careers at CWI, an increasing number of young staff members find employment in these sectors, for example in spin-off companies that are based on research results from CWI. Of course, library and computing facilities are first-rate. CWI's non-scientific services to its personnel include career planning, training and courses, and assistance in finding housing.

CWI has a vacancy for a PhD Student in the area of Evolutionary Computation, more specifically regarding Estimation-of-Distribution Algorithms, with applications in Energy Networks in the research group “Multi-agent and Adaptive Computation (SEN4)” in cooperation with Utrecht University (Computer Science).

PhD Student in Evolutionary Computation for Mixed Continuous-Discrete Optimization

CWI research group “Multi-agent and Adaptive Computation (SEN4)”, website:

http://www.cwi.nl/research-groups/multi-agent-and-adaptive-computation

Project description

Mixed continuous-discrete problems are hard non-linear optimization problems that occur in many application fields and are particularly difficult to solve. The variables of mixed type problems belong to different kinds of search spaces, such as continuous (or real-valued) variables, ordinal discrete (or integer-valued) variables, and nominal discrete variables, where the values have no order between them. The goal of this research project is to extend a class of meta-heuristic search algorithms - namely Estimation-of-Distribution Algorithms (EDAs), which are a type of Evolutionary Algorithms (EAs) – such that they can be applied to mixed continuous-discrete problems.

EDAs have shown great success in both the purely continuous domain as well as in the purely discrete domain. This project concerns the design of novel EDA approaches for solving mixed continuous-discrete problems. To ensure that the techniques that are developed are applicable to complex real world problems, a difficult but representative problem is targeted within this project, namely the Transmission Network Expansion Planning (TNEP) problem. This is an important problem in the real-world application area of future electricity networks.

The TNEP problem consists of defining when and where new circuits should be installed to serve, in an optimal way, the growing electric energy market, subject to a set of electrical, economic, financial, social and environmental constraints. Proper allocation and expansion decisions are essential for stable and efficient operation of the networks, with many durable suppliers.

This project is funded by NWO, The Netherlands Organisation for Scientific Research, and entails a collaboration between CWI and Utrecht University. The PhD student will therefore be supervised by scientists from both Utrecht University and CWI, but the student will be based at CWI.

Job description

Within this project it is your job to develop EDAs for solving mixed continuous-discrete optimization problems with a particular interest on solving the TNEP problem. Key tasks are combining existing EDAs for continuous and discrete problem variables in various manners, designing completely novel methods of tackling continuous and discrete problem variables simultaneously employing techniques from statistics and probability theory and the incorporation of domain knowledge on electricity networks in the form of constraints and local optimization. Moreover, the PhD student will be expected to be able to write their own computer code for these algorithms (in e.g. C, C++, Java, Matlab, Python, ...) and to perform experimental analyses (via e.g. batch processing).

Keywords are evolutionary algorithms, estimation-of-distribution algorithms, black box optimization, mixed continuous-discrete (or mixed-integer) optimization problems and the transmission network expansion problem.