Arne Andersson - Fredrik Ygge
The project ``Techniques for Distributed Markets, Negotiation, and Resource Allocation'' (or DIMALLOC), has been going on since the summer 1998.
The project is part of a cluster centered around multi-agent systems and market-oriented programming.
The people directly involved so far are
As described above, Arne Andersson and Fredrik Ygge has moved since the application was written. However, this has no negative effect on the project, rather the opposite. The planned contents of the project remains.
Master's thesis.
Abstract
The energy market in Sweden and other countries is facing major changes. The electricity market (and other
energy markets) has been deregulated, or is going to be deregulated. At the same time the development in the
field of microcomputers and computer communications has opened a new field of possibilities with
communication between the energy utility and customer (or equipment at the customers facility).
This change puts new demands on the energy utility and at the same time it opens up new possibilities.
There is an ongoing work in a project financed by NUTEK that among other things deals with the development of effective algorithms on resource allocation. Two algorithms have been presented that deal with resource allocation and controlling a group of loads that in some cases could only be set in discrete steps. The aim of the control algorithm is to level out the fluctuations of the total load of the group.
My task in this work has been to build an implementation of these two algorithms that works in a distributed environment. Furthermore, I have done some tests with the implementation. Tests that earlier only has been made with the load scheduling algorithm on its own and I have shown that given some prerequisites, i.e. results from functions that are not yet implemented, it is possible to control a real group of such discrete loads. The loads were electric heaters in an ordinary house.
The main result of my work is the implementation itself and that it shows that the task of resource allocation and load scheduling is manageable.
Submitted manuscript, 1999.
Abstract
We consider resource allocation with separable objective functions defined over subranges of the
integers. While it is well known that (the maximization version of) this problem can be solved
efficiently if the objective functions are concave, the general problem of resource allocation with
non-concave functions is difficult. In this article we show that for fairly well-shaped non-concave
objective functions, the optimal solution can be computed efficiently. Our main enabling ingredient is
an algorithm for aggregating two objective functions, where the cost depends on the complexity of the
two involved functions. As a measure of complexity of a function, we use the number of subintervals
that are convex or concave.
To be presented at
The Practical Application of Intelligent Agents and Multi-Agents.
Abstract
As a consequence of the deregulation of the electricity markets in many
countries, energy utilities have started to pay high interest to
providing services (e.g. supplying a comfortable indoor temperature)
rather than merely selling electricity (kWh). At the same time we
witness a tremendous technological development where vendors plan
products like "smart refrigerators". Providing energy services while
utilizing production and distribution efficiently using vast numbers of
"smart equipment" requires computationally efficient algorithms and
conceptually attractive metaphors.
In this paper we present a novel approach based on autonomous agents negotiating over electric power to manage large numbers of communicating loads. Its advantages compared to traditional methods are described. The approach is currently being implemented in a field test in a distribution area in the South-East of Sweden.
Further information can be retrieved from the project home page at
www.enersearch.se/research/ITE/DiMalloc
Uppsala, 1999-03-02
Anre Andersson
Professor