Load flow algorithm improvements

Artelys is specialized in mathematical optimization, quantitative decision-making, and scientific modeling. Relying on its high level of expertise in quantitative methods, Artelys’ consultants deliver efficient solutions to complex business problems. They provide services to numerous industries: Energy, Transportation, Telecommunications, Manufacturing, etc. Artelys is an international company with offices in France (Paris, Lyon, Nantes), Canada (Montréal), Belgium (Brussels), Spain (Madrid) and the USA (Chicago).   

Artelys offers several products and services, including software solutions (mathematical optimization software, business specific customized and custom solutions), studies, consulting, training, etc. In particular, Artelys consultants deliver operational solutions that perform security analysis of power grids, i.e. simulations that check that the power grid can be safely operated on the short term (e.g. one day ahead). It involves performing a massive number of simulations named load flows that evaluate the distribution of power flows on the grid from generation to loads. 

Load flows are usually based on a Newton Raphson algorithm to solve a non-linear system of equalities. This approach has two limits: 

• More complex electrical devices behaviors are modeled with outer loops calling several Newton Raphson algorithms. 

• When the algorithm does not converge, it comes with no proof of infeasibility and with scarce information about the divergence origin. 

By transforming the problem into an Optimal Power Flow (OPF), with the introduction of well-chosen slack and binary variables, it is possible to tackle those limits. 

The intern will design and implement two new models for the existing OPF solver using Artelys Knitro, the leading non-linear optimization solver.  

One model will tackle several devices behavior integrating them in the optimization problem with constraints and binary variables. 

Second model will use slack variables and objective function to give more robustness and help to understand non-convergences. 

The candidate must be in his/her last year of master’s degree in computer science and/or applied mathematics and/or power systems and/or operations research. 

Required skills: 

• Mathematical optimization 

• Programming in Java 

Valued skills: 

• Power Systems 

• Fluent in French