Renewables, electric vehicles and heat pumps are part of a fundamental shift affecting our energy landscape. These low carbon technologies are key to helping the UK make the transition to a net zero future. Their timely integration and effective use are heavily dependent on the monitoring and control of the last mile of our vast electricity network.
The last mile of a long journey
Electricity networks are the largest human-made system to distribute electricity to homes, schools and hospitals. National Grid connects large power stations to regional grid supply points managed by Regional Distribution Network Operators (DNOs). DNOs gradually step down the voltage depending on where the power is being supplied: that means extra-high voltage for large industrial loads, high voltage for large organisations, and low voltage for individual homes.
This low voltage network - the journey from the substation to the power socket in our home - is called the "last mile". It's often the trickiest and most expensive section of the network for DNOs to monitor and control. DNOs must keep the voltage within statutory limits, but also respond to demand and consumption. And they have to do all this while delivering a reduction in CO2 emissions and managing the cost to their customers.
To accommodate new low-carbon technologies like electric vehicles, heat pumps and photovoltaics without hugely expensive network investment, we need to forecast electricity usage and network behaviour. At the moment, many DNOs make these forecasts using historical load profiles and fixed network information about customer numbers, types and electricity use. This is often inflexible, inefficient and incompatible with modern low-carbon needs and usage patterns.
To enable DNOs to monitor and control their low-voltage networks in real time, without having to monitor each individual substation, we need to make the last mile truly visible using big-data analytics.
Network templates for accurate predictions
Working with Western Power Distribution (WPD) through Ofgem’s Low Carbon Network Fund, we set out to develop new network templates using a three-stage, load-profiling method of clustering, classification and scaling. These load profiles analytically characterise generation, networks and consumption.
To predict the daily peak load of low-voltage substations we used a contribution factor approach. For each template we developed the contribution factor by a cluster-wise weighted constrained regression. This considered the contribution made by different customer groups to substation peaks, improving the accuracy of peak load forecasts by 80% compared to previous generic templates.
Through this work we can visualise energy usage patterns of various locations, types and customer mixes. This means DNOs can effectively plan the last mile, and maximise the integration of low-carbon technologies within the existing low voltage network.