The ‘Connection Journeys Artificial Intelligence (AI)’ project aims to assess the opportunity for AI across the High Voltage (HV) and Extra High Voltage (EHV) connection journeys through the delivery of an AI Roadmap and Proof of Concept (PoC). The project will continue the drive towards greater digitalisation and help achieve the RIIO-ED2 strategic outcome of ‘connectability’ in the wider context of connection reform.
Benefits
Time Saving… One of the benefits of AI is the time-savings that it can bring. In assessing the opportunity for AI across the connections journey from pre-application, to application and offer, to energisation, we are looking at an opportunity to save company time across numerous teams and across all of NGED’s license areas.
In a study of 35,000 workers across 27 global markets, it was found that employees saved on average an hour a day from the use of AI in the workplace. This was found to be highest in the energy, utility and clean technology sectors where the average time saving rose to 6.25 hours per week. When this is applied to NGED’s staff working within the HV and EHV connections journey the time saving could be more than £2.5 million per annum. This was calculated by taking the salaries of staff who work on a part of the HV and EHV connection journey, assuming that 50% of their time is dedicated to HV and EHV connections, and applying the average hour a day AI saving found in the study above.
With this first phase of the project estimated to cost £418,859.10, this presents a strong initial case for investment. Time freed up would enable staff to focus on other tasks, increase accuracy and increase our efficiency as a business, which will ultimately benefit our customers. The Customer Journey Analysis (WP1) segment of the project will enable us to get a more accurate understanding of the scale of the time-saving benefit for future projects.
Learnings
Outcomes
The main outcomes of the first phase of Connection Journeys AI have been:
- 70 journey pain points. Nine of these have a relevant AI solution.
- A roadmap to develop the nine AI use cases.
- A proof of concept of AI use case focused on producing budget estimates.
During this first phase of Connection Journeys AI, an exploratory stage identified where AI can make substantial improvements to the process. This resulted in nine AI use cases which were subsequently prioritised and developed into a Roadmap to progress AI development and implementation through RIIO-ED2, ED3 and ED4. The final part of the first phase of the project was to develop one of the AI use cases into a PoC. In collaboration with stakeholders, AI use case 3 which focuses on Budget Estimates was selected. The scope was further refined to focus on the HV where there is a higher volume of data for the AI to be trained on than at the EHV level.
At the project initiation stage, it was anticipated that technology readiness level (TRL) would progress from TLR 2 to TLR3. TLR3 has been achieved successfully with a PoC which showed a promising degree of accuracy when predicting Budget Estimates. Multiple methods of prediction were evaluated but XGBoost demonstrated the highest performance achieving a R² of 0.72 for absolute cost prediction thereby accounting for a substantial proportion of observed variance and indicating that it is possible to achieve a strong indicative budget estimate from the model trained on formal quotes.
There is a clear opportunity to develop this further by refining the PoC model into an Minimum Viable Product (MVP). Additionally, as the Cable Routing (AI Use Case 4) is a key influencing factor in predicting the Budget Estimate, the next phase proposal focuses on developing these two use cases in tandem. The aspiration is therefore to create a Budget Estimate and Cable Routing tool which will assist 11kV planners in producing early-stage guidance for customers. Once the development to the model has been completed, it is proposed that we run a pilot with a small group of 11kV planners to test the model and its impact on teams. We expect that this next phase would progress the TRL from three to six with any further phase of the project moving to deployment.
Lessons Learnt
This project has improved our understanding of what connections digitalisation can look like for networks. By exploring the HV and EHV Connection Journeys, NGED has gained a better understanding of the pain points that exist as well as the opportunity that AI could bring. The intention is to develop these use cases further and funding has been requested for RIIO-ED3 development.
Throughout the project, key lessons were captured in the project learning log:
- The end-to-end HV and EHV connection journeys within distribution network operators covers a broad range of stakeholders. This meant it was challenging to have a workshop covering the entirety of the journey and to make it meaningful for all stakeholders. In person workshops were a successful way of mapping the connection journeys when broken down into different stages. Follow up online workshops were a good way to address any clarifications and validate responses.
- Organisational structure has a profound impact on process and ways of working. An example is that for the HV where 11kV planners are distributed across local offices practices may be more variable compared to the EHV where Primary Network Design engineers are centralised across the four license areas (some element of this could also be down to the differing nature of HV and EHV design).
- Although the focus of the project was on AI, the nature of running workshops to map the connection journeys meant that there was significant learning about where business as usual improvements are required. This learning was coded, shared with teams and implemented into RIIO-ED3 planning.
- Having a wide variety of stakeholders in prioritisation was beneficial to ensure that different business perspectives were captured.
- Substantial development across our IT&D teams meant that the dependencies of the AI use cases and the logical order of development was more challenging to capture than anticipated.
- Evolving business priorities means that whilst a roadmap is always a useful starting point, the roadmap can only be considered a snapshot and continuous stakeholder engagement is vital to ensure that plans stay meaningful and relevant to the business.
- In developing the proof of concept, identifying and selecting the relevant data posed some challenges. One of the ways that this was overcome was by having a session with both an 11kV planner and a Data Analyst. This shows the benefit of involving both the business users and the Data teams when securing data.
