Beta 1

Title Convex Optimization Algorithms for Power Plant Operation
Author Sokoler, Leo Emil
Supervisor Jørgensen, John Bagterp (Scientific Computing, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Bachelor thesis
Year 2009
Abstract DONG Energy is the main power generating company in Denmark. It operates a portfolio of power plants and wind turbine farms for electricity and district heating production. The wind turbines constitute a large share: 30 % of the installed generation capacity in Western Denmark. The share is expected to increase even further as a new wind turbine park is added to the portfolio at the end of 2009. In addition a large pool of electric cars are added to the power network. In a liberalized electricity market, such an interconnected power and heating system with significant stochastic generators and consumers needs an agile and robust control system. The control system is responsible for coordinating the most economic power generation respecting constraints, long-term contracts, and short-term demand-fluctuations. By simulation, Model Predictive Control (MPC) has been demonstrated as a very promising technology for dynamic regulation and coordination of power generation in the DONG Energy portfolio [1]. In MPC applications, the performance and reliability of the optimization algorithm solving the constrained optimal control problem is important, as the optimization problem is solved repeatedly on-line. Constrained optimal control problems for linear systems with linear constraints and an objective function consisting of linear and ℓ1-norm terms can be expressed as linear programs. The ℓ1-norm linear programs have a special structure that can be utilized. Using the special structure, we have developed an efficient primal-dual interior point algorithm “MPCIP” for such ℓ1-norm MPC problems. The primal-dual interior point algorithm is based on Mehrotra’s predictor-corrector algorithm [2] and efficient Schur-complement based linear algebra operations for this particular LP. The algorithm is implemented in Matlab and its performance have been compared to a general interior point algorithm, an active set based QP solver and linprog from Matlab’s optimization toolbox. Figure 1 demonstrates that the new algorithm is more than one magnitude faster than the other general purpose LP algorithms.
Imprint Technical University of Denmark (DTU) : Kgs. Lyngby, Denmark
Series IMM-B.Sc.-2009-20
Original PDF bac09_20.pdf (3.87 MB)
Admin Creation date: 2009-07-02    Update date: 2010-10-28    Source: dtu    ID: 246128    Original MXD