Are you using the right tools to solve business challen...
“In this ground-breaking project, we’ve used predictions, data, and expert analysis to learn so much more about how, where and when our customers might find themselves in more vulnerable situations. ”
The energy sector is undergoing enormous change driven by the rise of renewable generation, the move towards a net zero-carbon grid and escalating energy prices. As network operators make significant investments to upgrade their grids to support low carbon technologies, a vital question has arisen: how can we design a fair and just transition, making sure vulnerable customers aren’t disadvantaged or left behind?
In a pioneering collaboration funded by the Network Innovation Allowance, Scottish and Southern Electricity Networks (SSEN) joined forces with Smith Institute, Imperial College London and National Energy Action to develop a first-of-a-kind Future Energy Scenario – the Vulnerability Future Energy Scenario (VFES) – which helps operators understand what makes customers and communities more or less resilient, where vulnerability and fuel poverty are most prevalent and which factors will drive change in the coming years.
This case study details how Smith Institute helped SSEN use explainable AI to identify the drivers of energy vulnerability across the UK, supporting inclusive infrastructure planning in the face of energy transition.
In a pioneering collaboration funded by the Network Innovation Allowance, Scottish and Southern Electricity Networks (SSEN) joined forces with Smith Institute, Imperial College London and National Energy Action to develop a first-of-a-kind Future Energy Scenario – the Vulnerability Future Energy Scenario (VFES) – which helps operators understand what makes customers and communities more or less resilient, where vulnerability and fuel poverty are most prevalent and which factors will drive change in the coming years.
This case study details how Smith Institute helped SSEN use explainable AI to identify the drivers of energy vulnerability across the UK, supporting inclusive infrastructure planning in the face of energy transition.
DESIGNING AN EXPLAINABLE DATA-DRIVEN APPROACH TO PREDICT VULNERABILITY
The Challenge
SSEN is a forward-thinking distribution network operator (DNO) that maintains and operates the electricity distribution infrastructure for 3.8 million customers across the north of Scotland and central southern England.
SSEN leadership recognise the challenges presented by the transition to net-zero, particularly for vulnerable customers who rely heavily on secure, affordable and reliable electrical supply. Aware that their existing Distribution Future Energy Scenarios did not consider consumer vulnerability, SSEN aimed to support vulnerable customers and prevent new forms of vulnerability when making infrastructure and operational planning decisions. SSEN entrusted us with the task of harnessing their extensive data resources and employing machine learning to identify the factors driving energy vulnerability along with their impact.
The effort was complicated by the diversity of SSEN’s 3.5 million household customer base. Various driver groups interact in intricate ways to impact vulnerability, which cannot be simply attributed to a single factor like age. For example, within elderly customers, there are sub-populations with varying wealth, activity and health conditions which combine differently to impact vulnerability. We therefore needed to identify each of the groups that exist and determine how their demographic features drive vulnerability. It was also important to make sure that these findings were explainable and interpretable to decision-makers, making “black box” models unsuitable.
UTILISING EXPLAINABLE AI TO UNCOVER VULNERABILITY DRIVERS
The Solution
In our collaboration with SSEN, we developed an explainable AI solution to identify energy vulnerability drivers and their influence on different regions. Using advanced techniques to incorporate explainability into the model, we were able to capture complex, non-trivial relationships that exist in the data whilst retaining the ability to interpret what the model learns.
Vulnerability in Great Britain is measured at the household level and recorded on the Priority Services Register (PSR). Members of the public may join the PSR under a range of conditions that include reaching state pension age; having a disability; having a hearing, sight or mental health condition; or needing to use medical equipment that requires a power supply.
Our model used a range of demographic features including metrics related to population age, health benefits, income, social isolation, internet usage, housing and qualification to predict vulnerability in SSEN's serviced areas. It generated predictions and explanations for each area's vulnerability, detailing how factors combined to influence it. We then grouped areas with similar vulnerability drivers, applying data-driven clustering which helped us reflect natural patterns. To overcome high-dimensional data challenges, we reduced dimensionality before clustering, resulting in more representative groups.
Through this process, we identified key vulnerability drivers and presented insights into SSEN's diverse customer base. This gives SSEN better decision intelligence on the scale and location of such situations as well as what new situations may cause vulnerability.
“In this groundbreaking project, we’ve used predictions, data, and expert analysis to learn so much more about how, where and when our customers might find themselves in more vulnerable situations.
This level of understanding is already being put in place to plan our network investment strategies, not just in regions which will see a high uptake in low carbon technologies, but also communities where customers rely on energy more than most; and who may need more support with using low carbon solutions.”
This level of understanding is already being put in place to plan our network investment strategies, not just in regions which will see a high uptake in low carbon technologies, but also communities where customers rely on energy more than most; and who may need more support with using low carbon solutions.”
Simon O'Loughlin
VFES Project Manager
BUILDING DECISION INTELLIGENCE FOR A JUST AND INCLUSIVE ENERGY FUTURE
The Result
Our explainable AI solution provided SSEN with a detailed understanding of their most vulnerable customers and the factors driving their vulnerability.
Having identified the distinct groups of areas with similar drivers of vulnerability, SSEN can now take steps to ensure future investment strategies are designed appropriately to provide a just transition of the energy network for all. Localised and targeted investments and interventions can be made to address the specific drivers of vulnerability in each location, and strategies can be shared across different locations where similar drivers of vulnerability are dominant.
As a result of this ground-breaking project, SSEN is in the process of embedding Vulnerability Future Energy Scenarios (VFES) into its Distribution Future Energy Scenarios process – ensuring that the network takes vulnerable customers into account as it makes million and billion pound infrastructure investments for the future – ensuring that the net zero transition is also a just transition.
