Digital twin of energy distribution network.

As the UK strives towards to the goal of Net Zero by 2050, the electrical distribution network operators anticipate a rise in utilisation of Low Carbon Technology (LCT) such as electric vehicles (EV), heat pumps (HP) and photovoltaic solar (PV). An increase in LCT device connections to the UK distribution grid causes problems

with network power quality, network balancing and infrastructure investment. Load monitoring provides DNOs the data to make informed decision-making processes, however "intrusive" load monitoring is slow and expensive due to requiring accurate logs and details of device usage. Non-Intrusive Load Monitoring (NILM) is a

developed concept which has been proposed to provide a similar granularity of device power consumption, but has mainly been limited to household-level and uses "smart meter" measurements along with device measurements in the form of National datasets such as REDD etc. These measurements are fed into machine

learning algorithms to disaggregate the aggregate smart meter reading into sub-reading for the individual devices in the house. Whilst a great concept, it is difficult to efficiently apply to large geographical areas and hundreds of properties. Elliott's work will explore the use of NILM when applied to substation-level measurements to

disaggregate the constituent devices, along with applying advanced machine learning techniques to quantify the number of LCT devices at the substation based on a real-time measurement.

This is achieved using python-based machine learning libraries such as tensorflow and keras to implement bespoke supervised deep learning algorithms. Developed algorithms explore the use of artificial neural networks, specifically Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) with the respective subsets Long Short-Term Memory, Gated Recurrent units.

This work is purely data-analytics and machine learning based, with real-data being recorded and supplied by EA Technology ltd using their VisNet devices. Once this work is completed it can be used to understand the trajectory of the Net-Zero throughout the power distribution network, inform network investment and quantify the rate of adoption of Low-Carbon technologies.