(127c) A Semantic Representation of Policy Goals in the Modeling of Electricity Generation and Water Treatment Systems

Optimization models are used to evaluate different scenarios during the
formulation of energy and water policies, including those formulated to ensure
electricity and water supply security and sustainability.  The strength of these models lies in
their theoretical foundations built on mathematical equations that process
numerical (quantitative) data. 
Nevertheless, a complete consideration of energy and water policy issues
also requires the evaluation of non-numerical (qualitative) data.  For example, MARKAL, MESSAGE and DNE-21
are optimization tools that aid in the evaluation of energy policies, but their
use requires the intervention of modelers to "translate" the policy goals set
by the policy maker into sets of equations and constraints that are solvable by
mathematical procedures.

This paper introduces a modeling environment that can be used for policy analysis
and evaluation. This new modeling
system aims to support policy makers by automatically converting their policy
goals and targets as parts of an optimization model consisting of a set of
mathematical expressions. To
facilitate the inclusion of these additional aspects in a computer model, a modeling system, which operates at a level that is above that of optimization models, is proposed.

The modeling system integrates semantic representation techniques and knowledge
inference functionalities with an optimization model and its solver.  The selected method for semantic
representation is a formalism known as ontologies
which encodes knowledge in
a manner that can be communicated and shared between people and software
tools.  Ontologies
add semantics through a collection of associated terms, their conceptual
relations and a set of logical axioms. 
In addition, in this work we have integrated this semantic
representation to a suite of engineering models
(mathematical equations).

Case study: Prototype Electricity Generation Modeling System

The use of ontologies within
the area of energy modeling is a new development.  During the modeling task a modeler is required to translate the
intention of the policy maker, as defined by a set of policy goals and targets,
into mathematical equivalents fit for input into the electricity generation
model.

An added complication is that a comprehensive formulation of an
electricity generation mix must include aspects associated with the triple
bottom line sustainability (social, environmental and economic criteria), an
evaluation of which requires the consideration of a significant amount of
diverse non-numerical information.

Two prerequisites must be fulfilled for energy models to consider the
whole spectrum of sustainability aspects. 
First, the information associated with sustainability in the context of
energy policies must be identified and defined.  Second, a new approach to optimization-based energy modeling,
which considers both quantitative data and qualitative information, must be
developed.

We have developed a prototype system that uses a semantic representation
of energy policy knowledge containing not only quantitative data related to
technical, economics and environmental aspects, but also qualitative
information related to social and political issues.  Goals associated with sustainability issues are translated,
via the ontology, to equality and inequality constraints using description
logic.  Currently, goals can be set
with regards to the following criteria: capital cost, life-cycle energy
payback, CO₂ emissions reduction, land utilization, water
conservation, public health risk, new jobs creation, social acceptability and
energy supply security.  The
semantic representation is integrated into a prototype energy modeling system
used to formulate an optimized electricity generation mix.

A scenario-based analysis has been chosen to
incorporate the uncertainties associated with the future.  In addition, a bottom up approach was
chosen to represent technology from an engineering point of view.  Furthermore, optimization was selected as the most appropriate method for the
allocation of energy resources.

The Prototype Energy Modeling System uses a semantic
representation module which consists of four ontologies
with links to engineering models
(e.g. mass and energy balances). 
Representation of the core knowledge is formed by the Energy
Policy Ontology and the Value
Partition Ontology with
support from engineering models, while the Scenarios Ontology and Equation
Construction Ontology constitute the representation of the
application knowledge.

To demonstrate the Prototype Energy Modeling System, an example
case study for the formulation of an electricity generation mix in Malaysia was
conducted, which explores the influence of our extended set of policy goals and
targets on the optimization of electricity generation mixes, i.e. the selection
of type of electricity generation technologies and their quantity over a time
horizon.  The case study illustrates how through the
use of logic inference, the Prototype Energy Modeling
System
can formulate mathematical expressions to be used as an input to
the optimization of an electricity generation mix (the objective function, day
and night electricity generation requirements, renewable energy generation
limits due to intermittency, and heat and power requirements).  It demonstrates that for a set of
scenario drivers (population size and growth, industrialization levels, GDP
growth and decommissioning rates), an electricity generation mix that fulfils
the constraints can be obtained. 
There was a fair degree of similarity between the results generated from
the Energy Model created by the Prototype Energy Modeling
System
and the official Malaysian government plans, which lends credence to the
capability of the Prototype Energy Modeling System to formulate
electricity generation mixes that are acceptable and believable.

Case study: Prototype Water Modeling System

Similarly, optimization models aid in the formulation of water policies,
including those designed to ensure water supply security and
sustainability.

A prototype water modeling system based on a semantic representation of
water policy has also been developed. 
The objective of this second case study was to demonstrate the generality
of the overall architecture of the system: the Energy Ontology was substituted by a Water Ontology, the library of electricity generation engineering
models was substituted by its equivalent for water treatment and supply, and
the user interface was modified to reflect the new application (the new set of
goals).  The rest of the system,
i.e. the Knowledge Inference, Equation Builder and Linear Programming modules were not
modified.  As a result it took
approximately only four weeks to develop the Prototype Water Modeling System.

To demonstrate the concept of the second prototype, a case study for the
state of Penang in Malaysia was developed.  From a set of water policy goals, a water model was created
automatically and subsequently optimized to generate a water treatment and
supply system.

Discussion

         The
Optimization Modeling System presented
in this paper, addresses the disconnect between
the policy maker and the optimization model itself.  We have sought to extend the boundaries
of the formulation of optimization models used for policy evaluation and analysis
with particular emphasis given to energy and water models.  This has been achieved by semi-automating
part of the function and the associated tasks of the modeler.  This modeling system creates
a Model by automatically formulating the necessary mathematical
expressions for optimization based on both quantitative and qualitative inputs from
the user.  The modeling
system achieves this task through the combined use of a semantic representation,
knowledge inference, mathematical techniques and procedural programming.

         Our
modeling systems breaks new ground with regards to the
use of ontologies. 
As far as can be ascertained, this is the first use of ontologies as a method to formulate mathematical expressions
through inference in the knowledge domain.  This allows for a reduced dependence on
manual programming of mathematical expressions where human intervention results
in lack of uniformity and human error can cause inaccurate results.  From an application standpoint, the Prototype
Modeling Systems and the Models formulated offer the
advantage of a single evaluation platform, explicit model
documentation, rapid model reconfiguration, ease of use, and an opportunity
to explore new results.

         Our
modeling systems however do have limitations. In particular,
while the Optimization
Modeling System accepts qualitative inputs, the analysis of these
inputs using qualitative data remains limited.  From an application standpoint, the ability
of the Optimization Modeling System to process qualitative data is
limited by the richness of the ontology, which in turn is limited by the
available knowledge for codification.  In addition, the Prototype Modeling System and the
Model occasionally suffer from the vulnerability of an infeasible solution
space for which only limited problem finding and solving functions have been
incorporated.

         Our
proposed modeling system would benefit from additional research in the areas of
semantic representation and its integration to linear programming.  The four areas of improvement are:
(a) the expansion of the knowledge codification paths, (b) the improvement in
the integration between the ontology and the engineering models, (c) the
enhancement of the representation quality of the ontology, and (d) the adoption of a better
multiobjective function optimization
method.

         In
conclusion, we view the prototype modeling systems introduced in this paper as
a proof-of-concept showing their potential use by policy makers in their
efforts to formulate, analyze and evaluate sustainable and effective policies.