| Abstract
|
The increasing complexity of the smart gri … The increasing complexity of the smart grid raises concerns with performance, privacy,
security and dependability that go further beyond electrical network faults. In this
regard, electrical network self-healing and commercially available security solutions are
capable of handling a set of electrical network, systems and communications faults automatically,
but separately. However, as shown by the Ukrainian incidents, in 2015, there
can be cause-effect connections between faults and failures in different smart grid layers.
Additionally, although a set of European projects is addressing the security and dependability
of self-healing use cases, the pilot projects focus mainly on functional issues,
possibly compromising the security of future roll-outs.
We use a knowledge-based and security-by-design approach to design and propose a
secure and dependable Self-Healing System (SHS) with awareness of the aforementioned
connections. It is a Multi Agent System (MAS) with replicated Self-Healing Expert Entity
(SHEE) agents. Each SHEE is responsible for the self-healing process in a limited
domain, corresponding to a set of systems, components and processes assigned to its
scope of supervision. It reasons with knowledge based on facts and rules. It monitors the
domain, diagnoses eventual faults, creates recovery plans and reconfigures the smart grid
based on these plans. It cooperates with other SHEEs. It learns from the results and consequences
of its actions. It comprises a set of security and dependability features to prevent
and tolerate faults and intrusions, resulting from a threat and vulnerability assessment.
We perform a partial implementation of our system, consisting in the definition of
a self-healing domain, the corresponding ontology, the knowledge model with facts and
reasoning rules and a set of goals and queries. We successfully validate the SHS concept
as a solution to the described problems. The goals and queries are submitted to a
standalone inference engine, which is previously loaded with the knowledge model, simulating
the behavior of a SHEE replica through the different states of the self-healing
process. The process is repeated for four different complexity increasing fault and failure
scenarios. We discuss and provide guidance for a set of design and implementation issues
that, being critical to the security and robustness of the SHS, depend on each smart grid
specific context. epend on each smart grid
specific context.
|
