October 2022
Monday, 24th October
Speaker : Felix Agner, PhD student with Pauline Kergus at Lunde, Sweden
Title : Control to combat district heating bottlenecks: Distributed control for saturating actuators with nonlinear interconnections
Abstract : District heating systems form a large portion of the domestic heating sector in Sweden. The traditional architecture to control these systems is that central production units heat up water, and use pumps to create a large pressure to drive the hot water through pipelines to customers. The customers use local regulation valves to decide how much hot water they shall receive. These valves are normally controlled in a completely distributed fashion, and under normal conditions that works rather satisfactory. However, under certain conditions the traditional system breaks. When all customers want too much hot water at the same time, the pressure generated by the pumps is not enough. Buildings close to the pressure source can fully control their flows and stay warm, but units far away will experience a drop in temperature. The control problem I am looking to solve is to design a controller that (i) Is fully distributed under normal conditions, (ii) Coordinates the units to a fair distribution of heat when the disturbances on the systems are too great.
In this seminar I will present a quick overview of the mathematical model which captures the constraints of the system, then I will show a proof-of-concept controller that solves the problem, and then talk about ongoing work where a generalized linear simplification of the problem is solved optimally by distributed PI-controllers and a global anti-windup scheme.
Title : Control to combat district heating bottlenecks: Distributed control for saturating actuators with nonlinear interconnections
Abstract : District heating systems form a large portion of the domestic heating sector in Sweden. The traditional architecture to control these systems is that central production units heat up water, and use pumps to create a large pressure to drive the hot water through pipelines to customers. The customers use local regulation valves to decide how much hot water they shall receive. These valves are normally controlled in a completely distributed fashion, and under normal conditions that works rather satisfactory. However, under certain conditions the traditional system breaks. When all customers want too much hot water at the same time, the pressure generated by the pumps is not enough. Buildings close to the pressure source can fully control their flows and stay warm, but units far away will experience a drop in temperature. The control problem I am looking to solve is to design a controller that (i) Is fully distributed under normal conditions, (ii) Coordinates the units to a fair distribution of heat when the disturbances on the systems are too great.
In this seminar I will present a quick overview of the mathematical model which captures the constraints of the system, then I will show a proof-of-concept controller that solves the problem, and then talk about ongoing work where a generalized linear simplification of the problem is solved optimally by distributed PI-controllers and a global anti-windup scheme.
Speaker : Corbinian Schlosser, POP Phd Student
Title : Conditional expectation estimators using moments
Abstract : The presented approach was motivated by state estimation for dynamical systems. This task describes the problem of estimating a state of a dynamical system from partial and noisy observations. I will shortly introduce how certain instances of this problem can be embedded into the language of occupation measures for dynamical systems combined. This perspective leaves us with the problem of estimating the conditional expectation of
given 
from moments of the underlying joint distribution 
. I will present our method which provides a convergent series of estimators for the conditional expectation relaying only linearly on the moments.
Title : Conditional expectation estimators using moments
Abstract : The presented approach was motivated by state estimation for dynamical systems. This task describes the problem of estimating a state of a dynamical system from partial and noisy observations. I will shortly introduce how certain instances of this problem can be embedded into the language of occupation measures for dynamical systems combined. This perspective leaves us with the problem of estimating the conditional expectation of
given from moments of the underlying joint distribution . I will present our method which provides a convergent series of estimators for the conditional expectation relaying only linearly on the moments.