Part II Socio-Economic Modelling
Chapter 5 Improvements in the Representation of Behaviour in Integrated Energy and Transport
5.6 Conclusion and Recommendations
This chapter analysed twenty-seven integrated energy and transport models and created a taxonomy for these various model types.65This chapter reviewed the methodologies adopted for introducing behavioural features related to consumer purchase, adoption and use of transport technologies with the purpose of addressing two questions: (i) how should transport and energy models be structured to allow an effective inclusion of behaviour and (ii) what key attributes and parameters should be introduced to represent transport-related consumer choices in an integrated energy and transport model. Relating to the former question, the chapter concludes that there are three common approaches for structuring a model to improve the representation of behaviour - top- down, bottom-up, and hybrid structures - each of which have advantages and disadvantages
depending on the scope and purpose of the model analysis. Nonetheless, soft-linking and novel approaches recently developed (see Section 5.4) emphasise a bottom-up model structure as the most flexible and promising method. Concerning the latter question, this review identified technology choice, modal choice, driving patterns and new mobility trends as the key features to correctly depict transport behaviour in integrated energy and transport models (E+T). Furthermore, heterogeneity, travel time budgets and driving profiles are recommended as the key attributes and parameters to be introduced in ‘E+T’ models to represent such behavioural features.
5.5.1 Structure
Top-down (TD) models examine the entire economic system in a detailed way and constitute a valid
tool to simulate the economic mechanisms that regulate technology substitution. They can be used to endogenize modal or vehicle choice and to answer research questions concerning the relationship between modal/technological demands and fuel/electricity prices. Nonetheless, having the
economic sector as the core and focus of the model, TD models may fail at including a
comprehensive set of fuels, vehicle technologies and modes. The attributes characterising the different transport alternatives are often rendered to a low level of accuracy and TD models are less capable at directly capturing the effect of changes in efficiency, mileage and occupancy factors, relative to BU models. Further efforts are required to bring such models to a technologically rich format, as done by Karplus et al. (2013).
6 Sectoral energy models (E), energy models partially including the transport sector (E+), highly integrated energy and
transport system models (E+T), transport models partially including the energy sector (T+), and sectoral transport models (T) (see Section 5.3)
On the contrary, bottom-up (BU) models more suitably analyse the effect on the overall energy system of certain exogenously imposed modal/technology shares. As long as vehicle market shares are exogenous assumptions, pure BU models prove to be a valid policy analysis tool. Inversely, to endogenously determine behaviourally realistic market shares, the BU framework needs to be upgraded by adding new variables or by linking it with external socio-economic transport-focused models, of the ‘T’ or ‘T+’ types.
Hybrid models join and harness the advantages of BU and TD frameworks, thus proving more
capable at capturing many of the behavioural features discussed. They are valuable at answering research questions investigating both the energy sector and the surrounding economy. However, the structure of this class of models is inherently more complex when compared to pure BU or TD models, potentially creating issues with computation. The model comes as a “single-package”, not separable into the TD and BU components, thus limiting the flexibility of its use.
Of the three structures outlined above, this review regards BU models as the most promising approach to include a representation of behaviour in E+T models. The benefit of representing behaviourally realistic choices directly within an energy system model is manifold. Firstly, these improved BU models allow for energy system-wide considerations. Secondly, they support in the understanding of the future reciprocal implications of decisions taken in the transport and energy systems. Thirdly, a much wider variety of policies can be assessed through the E+T framework, as further discussed in Section 5.6. Because BU optimization models do not originally represent
behaviour, either they need to be soft-linked with an external transport simulation model which has a predefined representation of behaviour, and uses a complementary mathematical method, or their structure needs to be adjusted to accommodate the new behavioural features. The former approach makes the model flexible - whenever the analysis is not purely transport focused, the energy system model can run in standalone mode with a simplified representation of the transport sector. The latter approach is further discussed in Section 5.5.2.
5.5.2 Parameterisation
Technology and behaviour measures have been identified as critical measures in addressing
transport CO2 emissions, in particular, avoiding, shifting, and improving (IEA, 2012a). For this reason, this chapter aimed at identifying the most suitable method(s) of representing technology choice (improving), modal shifting (shifting) and both driving patterns and new mobility trends (avoiding) in ‘E+T’ models.
Including heterogeneity was regarded as the best means of improving the representation of
transport technology choice. Traditional BU energy system models assume homogeneous consumers
taking perfectly rational decisions. Introducing heterogeneous decision makers is a precondition for incorporating behaviour in ‘E+T’ models. Heterogeneous transport users have different preferences, resulting in a wide portfolio of technologies chosen, each one optimal for a specific consumer group. When deciding the number of dimensions along which consumers are split and the number of behavioural features to consider, a compromise between model complexity and completeness needs to be made. An ideal representation of transport behaviour within an ‘E+T’ model would involve representing all consumers within the region in question, yet the computation power required for this level of detail renders this method incredibly onerous. To avoid intractability or excessive complexity of the model, efforts should be addressed towards determining the minimum number of dimensions and subgroups necessary and sufficient to distinguish the main consumer groups in an exhaustive way.
Of all approaches reviewed regarding modal choice representation, travel time budget (TTB) is recommended as the best method of modelling this feature within BU models. It can be introduced by adopting literature values (Schäfer and Victor, 2000) or eventually more region-specific TTBs, available from national travel surveys. Moreover, the concept can be easily incorporated in the model, requiring only the definition of modal speed and the setting of a constraint (as in Daly et al. (2014) and Pye and Daly (2015)). An interesting area for future work would be to adapt the
methodology proposed by McCollum et al. (2016) and Bunch et al. (2015) to cover modal choice in BU optimization models – to provide reliable modal shares and calibrate the intangible costs
suitably, the energy system model requires drawing data from a detailed support model (e.g., of the ‘T’ type) that incorporates modal choice.
There is a need to model driving patterns at a detailed geographical level to accurately account for fuel consumption and emission factors from vehicles, which strongly depend on the driving performances. The relationship between modal speed and infrastructure could be incorporated in the integrated energy and transport model as was carried out in the model IMACLIM-R (Waisman et al., 2013). Another possible method consists in adapting the approach by Ramea et al. (2016), where the congestion level, and thus the modal speed and emission factors, is determined in an iterative way as a function of the infrastructure capacity.
Modelling new mobility trends offers the opportunity to explore and unlock their potential in
contributing to more sustainable transportation systems (Wadud et al., 2016, Grischkat et al., 2014). Car sharing services can possibly be modelled by introducing a new car technology type
characterized by a higher mileage per year. Carpooling can be incorporated in integrated energy and transport models by considering a lower car-ownership level and higher occupancy factors.