Why we use models

We use models to answer questions about important energy issues that affect UK and global societies.

Example of the utility and limits of models

All integrated assessment models (e.g. TIAM-UCL-IAM) tend to use economics as the main criteria for decision-making and aim to minimise total energy system cost.  This means that scenarios from these models tend towards normative descriptions of the future that simulate what should happen from an economic perspective rather than what will happen.

Important influences on the energy system, e.g. energy security, equity and affordability concerns, or often not included in these models but are at the forefront of politicians minds.  This is important both within and between countries.  For example, global models will find the least overall cost but this might create winners and losers in different regions of the world and those people who stand to lose out will resist the changes identified by the model.

The IPCC characterise integrated assessment models in terms of:

  • Economic coverage
  • Foresight (perfect or myopic)
  • Representation of trade between world regions
  • Assumed flexibility of global economies
  • Model detail (e.g. level of disaggregation of regions, technologies, etc.)
  • Representation of technological change

It is more appropriate to describe other types of models using different criteria.

Our approach to modelling

There will never be a universal model or method of modelling which will answer all questions.  We design our models to answer specific research questions, although some of our more complex models can also contribute to research in a number of different research areas.

In our experience:

  • Theory, models and practice are not the same (except in theory). We very often need a range of models for any given problem, including simple models that can run on a human brain.
  • The most important bits of a model are the people who create and run the model and use its output - developing the community of users and developers is as important as developing the software.
  • Models of complex systems grow and evolve best through structured contact with reality.
  • Models are only as good as the data you have to populate & challenge them. However, models driven with synthetic data can still shed light on problems.

We always remember that models are abstract representations of reality.  George Box commented that "All models are wrong, but some are useful."  It is important to be transparent about the design and inherent assumptions of models that we use; in fact, this is the aim of this website.  All models have important assumptions and limitations and even models with the same paradigm are often designed quite differently, as illustrated in the example of model utility and limits in the box opposite.

Our models

There are many different types of models.

Some of our researchers produce models (using, for example, statistical functions, cluster analysis, Bayesian models, etc.) for which the parameters or coefficients are estimated from measurements.  Such models can range from simple single-equation relationships to complicated econometric models.  This is often the most appropriate type of model for research into human behaviour, for example in the People and Energy research theme at the UCL Energy Institute.

Our researchers also build large models to examine research questions for which estimated models are not appropriate (for example, to understand the interactions between different energy systems with different drivers).  While these models are initially designed to address particular research questions, they are often used to contribute to other research questions beyond the context for which they were first constructed.

The models documented on this website generally fall into the latter category.  As such, they represent only a subset of the tools used at the UCL Energy Institute, which includes a range of modelling, data analysis and monitoring.

Our skills

We have experience of optimisation, simulation and statistical models, including both bottom-up and top-down models.  Types of model that we use include:

  • energy system models;
  • macroeconomic models;
  • econometric models;
  • stock models;
  • Bayesian belief models;
  • physics-based models; and,
  • agent-based models.

Members of the UCL Energy Institute use models written in many computer languages including Microsoft VBA in Excel, Python, GAMS, MATLAB, Fortran, etc.