Learnings

Outcomes

For 2016-19 and 2019-20, the forecasts provided by Innowatts, showed minor deviations from SSEN’s forecasts. It must be highlighted that Innowatts’ 2019-20 forecasts were aggregated from half-hourly figures into a monthly total, in order to make it comparable with SSEN’s data. This resulted in a comparison of two datasets of unequal granularity, hence not a clear evaluation of that year’s forecast was possible. The half-hourly forecasts produced by Innowatts provided a level of granularity which was not possible to evaluate and certainly highlighted the need for SSEN to collect more detailed information much closer to the customer interface.

Due to covid19 restrictions, the face-to-face dissemination workshop was converted to a virtually hosted session with participation of representatives from the other GB DNOs.

The original aim was to move the TRL of the method from 8 to 9. However, as the solution will not be progressed any further by SSEN at this stage, no work has been done in order to adopt the technology as a Business as Usual operation.

Lessons Learnt

The project produced the following learnings:
• It became apparent that Innowatts’ forecasting process capabilities exceeded the SSEN’s internal forecasting processes that are currently in place. The data granularity that Innowatts required included data such as the exact number of solar panels, or LCTrelated consumption. SSEN do not presently hold that level of detail, therefore, a more extensive evaluation of the forecasts was not possible by SSEN. This issue will be resolved once the roll-out of the second-generation smart meters has reached a suitable level of penetration. As soon as such data becomes available to the DNOs, the technology could be trialled further and potentially adopted on a larger scale. However, further deployment and adoption of the technology will not take place at this stage.
• The area of ‘fiscal’ forecasting and terminology that is used in the relevant documentation is considered niche. A sufficient level of knowledge and familiarisation is required in order for external parties to be able to interpret the data correctly. Therefore, it would be advantageous for the energy industry to produce a guidance document or reference material to streamline the approach and help other suppliers fully understand and accurately analyse the data. In this project, Innowatts managed to understand the data provided by SSEN, after a few initial sessions between the two parties.
• When the fields of forecasting and ML are combined, the trial highlighted the importance of providing the largest volume of data when possible. Historical data is key for ML models to learn, improve and produce more accurate results. Similar projects in the future need to be able utilise as much input data as possible for better results.
• The General Data Protection Regulation (GDPR) rules have been considered throughout this trial. As some of the data held by SSEN involved sensitive information by default, such as Meter Point Administration Numbers (MPANs), which are unique ID numbers associated with the locations of the customers’ meters. To protect the customers’ sensitive information, the SSEN’s data protection team were consulted and all the relevant data masked appropriately before data sharing took place. Such actions are also advised for any projects that handle similar data, so that compliance with GDPR is ensured.