Data is being collected and generated at a rate never seen before in human history and this is no different in the energy sector. Despite the frightening volume of data, it can quite often be difficult to find what you need or to find the connectives in the data sources that form representative structures of the domain.
To combat this, considerable effort has been placed into various technologies, such as, Big Data, machine learning and search engines that will attempt to discover the relations amongst data elements in the system from the data itself, to form networks and graphs; or in web-ontologies and linked-data that have languages that will describe each individual relation.
However, the data in many of the sources in the energy domain show that much of it is described through levels of abstraction. Where parts go to make up wholes, which in turn may form parts of other wholes, and where more than two elements are needed to form a semantic relation, capturing the thing being described. For example, a substation is made up of transformers, in-feeders, lines and many other parts, but the substation itself forms a part of the distribution or other power network. The diadic relations of networks provide a powerful way of representing this structure. However the relations that make up many of them are multiadic (n-ary rather than binary) and require a generalisation of network theory. Hypergraphs and simplicial complexes pave the way towards this generalisation which is provided by hypernetworks. The essential first step is the definition of a hypersimplex as an ordered list of parts (simplex) with an assembly relation that gives the information needed to assemble the parts into a whole at a higher level in a multilevel representation of the system. This algebraic structure is required to support the bottom-up top down dynamics of the system which is often but not always modelled by numbers representing local and global system states and transformations. In complex systems in general and energy systems in particular a part is not constrained to a single multi-level structure: a part can belong to many such structures forming an overlap where interaction amongst systems and subsystems can occur. For example the Air-source Heat Pump exists in both the power and heat networks as well as, it could be argued, the weather system.
Looking at the technologies often used to describe such systems, such as, XML, JSON, RDF, UML and others, all have been used to describe parts of the system, but on closer consideration, they all naturally form multilevel and n-ary relations; relations that do not form from raw data.
In an attempt to bring together these structures in a way that will facilitate mass connectivity and discovery in the energy sector, Hypernetwork Theory (HT) has been adopted as the modelling language of choice to create a single semantic schema. A set of software tools, including the HnC (Hypernetwork Compiler) have been developed to support the creation and manipulation of Hypernetworks. These enable a simple description of the system to be turned into a hypernetwork (Hn), or to merge them to create larger models or to facilitate different views to be created out of the model.
Not only does HT allow for the description of complex systems and facilitate the interplay amongst them through overlapping and other meronymic and taxonomic techniques, it also enables ontologies and other structures to be both ingested into the model and to be generated easily. HT makes modelling much simpler. Although web-ontologies with RDF, OWL and Description Logics create good models, they can be very difficult to develop and maintain, as each individual relation has to be individually defined, and even then the true semantics can be lost as the model does not capture the n-ary relation that actually forms the meaning. HT simplifies this by grouping the elements into a single n-ary relation where required, but still allows other relations amongst them if desired.
HT is flexible and adaptable, and given the tools available different models can be merged dynamically at runtime, and new ones created that can form new and important relations from the models, even allowing for models to be combined across domains.
This presentation will (i) review the theoretic basis of hypernetwork theory to create and implement dynamic computational models of complex evolving systems, (ii) illustrate this new theory by sketching a large industrial application in the energy sector, and (iii) discuss further developments of this approach and the implications for the development of large complex computational systems processing large quantities of heterogeneous data in real time.