The ambitious climate targets set by governments and the interconnection between energy and economic objectives requires careful planning and assessment. Among other instruments, energy and climate specialists often rely on models to inform their policy recommendations. Specialists have extensively used whole energy system models, like TIMES, and energy-economic models, such as Computable General Equilibrium (CGE). Each type of model presents different features and has strengths and limitations. However, the interaction of models, mainly using a ‘soft-linking’ approach, is recognised as a potential solution to overcome some of the drawbacks of the standalone models, and to improve the depth of analyses. The approach consists of using information from one model to inform the other in iteration until the two models are harmonised.
The University of Strathclyde Centre for Energy Policy and Fraser of Allander Institute are currently investigating best practices and challenges on the use and links of TIMES/CGE models for the analysis of future climate and energy scenarios in Scotland. The research is supported by ClimateXChange. Preliminary results from the research have been discussed during a one-day workshop in May 2018.
The event brought together stakeholders and researchers from different countries, to discuss practical challenges and share experiences. Presenters in the workshop included: Andrew Mortimer (Scottish Government), Patricia Fortes (New University Lisbon), Matthew Winning (UCL), Anna Krook-Riekkola (Luleå University of Technology), James Glynn (UCC), Anna Darmani (InnoEnergy), and the two organisers of the event Gioele Figus and Christian Calvillo (University of Strathclyde).
Gioele Figus and Christian Calvillo presenting their research
Each presentation provided a valuable point of view in the debate. Although, Portugal, Sweden, Ireland or Scotland may have different policy objectives, the presentations have highlighted common methodological practices and challenges that are applicable across different countries and regions. The process of linking the two models and use them in an iterative fashion raises a number of challenges that the Strathclyde team has summarised as follows:
- Model calibration, updating and data problems. Using TIMES and CGE is a dynamic process that requires constant updating and development. This requires considerable efforts and time. Moreover, the data to calibrate and update the models is not always available, and this further complicates the process.
- Validation of model improvements. When models are developed to improve the analysis of specific policies (e.g. modifying the CGE model to better represent the power system), it is difficult to check the reliability of the outputs. This is because the data might not be available to validate the changes made.
- Fundamental differences in the models’ objectives. The two models have been created with different objectives. TIMES chooses the technology mix that minimises the cost of delivering energy to a specific country or region. The CGE considers firms within the same region that produce goods and services by maximising their profit and sells them to consumers. To reconcile both objectives is not always an easy task.
- The soft-linking methodology is not straightforward. What to link and how? Which one is the dominant model? Where to start? Convergence criteria? These all are important questions that do not have a single general solution, as this depends on the application and analysis objectives.
- Data harmonisation for soft-linking. Mapping sectors and variables from one model to the other is not straightforward and it is likely to be different in both directions. For instance, the two models express energy in two different units of measure, monetary versus physical units.
- Sector bias on models. Researchers often focus on improving one single aspect of the two models, such as the treatment of heat or a particular industry. However, this could create imbalances in the solutions and potential reliability of the model, when a holistic approach is not considered.
- When to stop? Further improvements are always possible, but it is not always clear when new changes in the models are adding value to the analysis, making worth the effort, and when changes could bias the solutions.
There are clearly many challenges in the use of these two modelling families, and the workshop participants agreed to continue to collaborate and share information. However, some best practices have already emerged. These are:
- The importance of the baseline scenario – Each model considers a baseline against which counterfactual scenarios can be assessed. The consistency of the baseline in the two models is fundamental for a successful soft-link.
- Soft-linking is not always the solution. Some questions can be answered directly with one model or the other, removing the need of complicated soft-linking processes. Here, it is fundamental to have specialists with enough knowledge of the two models to assess which instrument is the most appropriate.
- Clear time frames for policy decisions help choosing the adequate model. As discussed previously, different models have been developed with different objectives and time horizons. For instance, TIMES can be used to answer how the energy system will look like in 30 years time, but it might not be the best tool to assess the changes in the energy system in next one to two years. It is, therefore, important to understand this in order to use the best tool for the problem at hand.
The main point to take home from this workshop is that an active and cohesive community of experts is a prerequisite for the correct and informed use of complex modelling technique for policy analysis. The workshop participants have agreed to increase the level of engagement by setting up forums and shared repositories of documents. Moreover, a second meeting will take place this time next year to continue discussing the progress made by the Strathclyde team and by the rest of the community.