# Self-Consumption Optimization of a Single-Family Home

Author(s):D. Caraci, Clemens Fischer |
Conference/Journal:Master Thesis, HS13 (10318/19) |

Abstract:In this thesis we propose a Model Predictive Control (MPC)algorithm which optimizes the self-consumption of renewable eneergy produced in a building system. We modeled a Single-Family Home (SFH) using standard Simulink components of Siemes. The modeled SFH uses renewable energy sources such as a thermal solar collector, a photovoltaic plant, and a heat pump with a ground as reservoir. Furthermore the SFH was equipped with two thermal storage tanks for heating and domestic hot water. The SFH was disigned for a heat demand of 45kWh/m2a which corresponds to a typical renovated building. The nonlinear SFH model was simplified to a Linear Parameer Varying (LPV) model having continuous as well as binary state and input variables. The parameters are predictable disturbances such as the outside temperature, the solar irradiation, etc. To reduce the complexity of the resulting optimization problem, the SFH model was split into an LPV subsystem of the building with only continuous variables and an LPV hybrid subsystem of the primary components (i.e. thermal solar collector, storage tanks, etc.) with continuous/binary variables. This split has the advantage that the Constrainded Finite-Time Optimal Control (CFTOC) problem of the building subsystem can be efficiently solved using a Linear Program (LP) solver. The CFTOC problem of the primary subsystem was formulated as a piecewise affine system and solved using a Mixed-Integer Linear Program (MILP) solver. Ther perfomance of the MPC was compared to a Rule-Based Control (RBC)formulation. The results of a one year simulation in a MATLAB/Simulink set-up showed, that the MPC ends up with higher electricity costs over the whole year. Likely reasons for the higher costs are that the chosen linearization oft the heat pump model is not accurate enough and the prediction horzizon is not sufficently large to predict the energy demands over a whole day. Our results led to the conclusion that the RBC approximates the optimal control well. | Year:2014 |

Type of Publication:(12)Diploma/Master Thesis | |

Supervisor:D. Sturzenegger | |

No Files for download available. | |

% Autogenerated BibTeX entry @PhdThesis { CarFis:2014:IFA_4788 } | |

Permanent link |