Information Technology Reference
In-Depth Information
determine the most appropriate organizational structure at run-time in the absence of
a central controller and in a scalable manner.
When multiple objections are being adressed to one coalition, its decision of consid-
ering one would be based on the criteria of maximizing parameter
τ
, while minimizing
μ
parameter
. The distribution of PA agents amongst VPPs is repeated until there are no
neighbor VPPs that would gain a higher utility value in terms of the interaction scheme
described above. Also worth remembering is that the procedure is ought to occur with
the domain dependent constraints, that impose maintaining the profile of the coalition
within certain limits (see Section 3). Finally, we stress that the aim for our proposed
scheme is intended towards an open-ended organizational adaptation concerned with
achieving stable configurations in dynamic environments where one-shot optimization
procedure are unapplicable.
5
Experimental Results
In this section we focus on emphasizing the results attained during the inter-coalition
interaction phase, based upon the solution concept introduced and projected on arbitrary
grid configurations. We use hereafter the notion of a coalition interchangingly with the
term VPP
3
. We simulate a system comprising of up to 1000 coalitions deployed in ar-
bitrary topological configurations. The generated configurations correspond to generic
meshed suburban network models, being identified as the most suitable setting for the
deployment of small-scale VPPs. This entails that each resulting VPP is considered to
be directly connected to at most 6 neighboring ones. Thus, we are referring to coali-
tional graphs limited to a 6-node connectivity, as depicted by a plausible VPP scenario
for this particular type of topologies. For generating the random graph structure we have
used the model proposed in [5]. The energetic capacity for the virtual power plants is
assumed at an average of 6MW (small-scale VPP, consisting of at most 500 DERs per
VPP). The simulations assume daily variations for the generated energy of each coali-
tion, bounded to an extent of at most 20%. For these experiments we have considered
commercially available residential DERs of 3 capacity classes of at most 25kw. We pre-
sume the distribution of DERs in the grid capable of matching the overall consumers'
demand. The results presented have been obtained by averaging over 20 realisations
(statistically significant for reducing to very low levels the results' variance).
To begin with, we first evaluate the performance of our algorithm attained through
the argumentation scheme introduced. Given the cooperative scenario reflected by our
chosen solution concept we have set aside from the Pareto optimal instance
4
where self-
interested agents agree to participate in a trade if and only if the contract increases the
agent's immediate payoff. This basic type of negotiations alone (unrestricted by topo-
logical configurations), where DERs would have individually been transferred between
coalitions, has proved to reach a local optima, with higher social welfare than others [1].
Alternatively, our chosen scheme for negotiation is primarily aimed at increasing
the social welfare of the system and thus, avoiding some of the imbalances that could
3
By VPP we divert from the common understanding per se and rather denote an ensemble of
both
LA
sand
PA
s, instead of merely
PA
s.
4
Represented in the graphs as the individualistic approach.
