Peak demand is subject to a range of uncertainties, such as population growth, calendar effects, changing technology, economic conditions, prevailing weather conditions (and their timing), as well as the general randomness inherent in individual usage. To improve peak forecasting, the relationship between peak demand and its driving factors must be understood across the short, medium, and long-term ranges (5yr, 10yr, 30yr).
The project will study the latest advancements in peak demand forecasting, both in GB and select regions globally, comparing against National Grid ESOS's (NGESO) current methodology. The second phase will focus on quantitatively assessing the drivers of peak electricity demand.
Benefits
- Findings from the project can be used to explain peak demand in its current state and the assumptions made from its future state to our stakeholders
- The project will improve understanding of uncertainty in NGESO forecasting methods and how to improve by targeting high return areas of improvement
- Evidence provided can support an industry discussion about risk appetite – e.g., how can stakeholders benefit from quantified uncertainty in future peak forecasts
- The project will help identify how societal behaviours that affect peak demand may change and under what conditions
Learnings
Outcomes
The project produced an initial model along with accompanying documents outlining the research findings, build method, model performance and suggested areas for future development.
The two reports listed below are published on the Smarter Networks Portal and learnings have been incorporated into FES 2023:
- Peak demand forecasting methodology
- Historical analysis of peak electricity
Lessons Learnt
Working arrangements
The project was undertaken during the latter stages of the pandemic and as such the meetings were carried out remotely. On the whole this worked well, but the project would have benefitted from some in person workshops, to collaborate on some spontaneous idea generation and discussion.
Methodology
Electricity demand is currently going through a period of change, with the movement towards net zero. Aurora have suggested a base regression model for demand forecasting. This has been constructed using traditional drivers based on their literature review: Temperature, day factors (weekend, weekday, bank holidays…). The model performed well during their tests, particularly in the near term. However, incorporating new demand behaviours such as: electrification of heating and transportation, requires alternative modelling methods and this will become increasingly important as the forecast covers deeper into the future. In the second phase of their work Aurora have considered these factors and proposed a hybrid model, utilising the original regression, along with individual demand elements that were derived from small datasets. From the methodology, the key learning point is that there is much more to do, the project has identified that many of the items highlighted in the list below are almost mini projects in there own right. Key areas within the method requiring additional understanding based on learning from the project:
- Further development of the regression model and its drivers.
- Analysis of the relationships between the individual drivers.
- Individual studies on new areas of demand and their effect on the peak forecast:
- Electrified heat;
- Electrified transportation;
- Hydrogen production;
- Behavioural changes (residential, industrial, commercial);
- Timing of peak for different demand groups.
- Develop an optimal hybrid technique for compiling the individual demand methods.
- Consider a bottom up modelling technique.
- Investigate the effect of alternative peak periods, not just winter.
- Incorporate global climate models.
- Introduce price forecasting elements.
- Carry out inflection point analysis.
Data
One of the key challenges faced by the team was the acquisition / availability of high-quality high-resolution data. This was particularly noticeable within new peak demand drivers, as little, or no data was currently available. The learning suggested that creating relationships with third party providers, such as academic institutions, or research institutes, could provide major benefits to further studies.
