Issue link: https://fhpublishing.uberflip.com/i/1173186
PARTNERSHIP STUDY On the generation side, re- newables are being created in a range of previously unexpected places. "This doesn't come into the network in centralised locations it comes in at every level, even from homes. Again the network was not designed for these levels of voltage or for the multi- direction flow. We need a smart grid to allow us to manage these generation and demand chal - lenges." Brazier says understanding data is critical to show where power is being generated and used, and also to highlight the condition of the assets distribut- ing that power. He points to the Low Carbon Technology Detection Project run by Western Power Distribu - ENA has set up its own data working group to try to maxim- ise networks' use of machine learning. "If we can use the technol- The role of AI John Moriarty, professor of mathematics at Queen Mary University and strategic lead for energy at The Alan Turing Institute, looks at the role for artificial intelligence in the energy sector. Areas of focus in artificial intelligence (AI) currently include reasoning, knowledge representation, automated planning and scheduling, machine learning, natural language processing, robotics and perception. But will these capabilities find an increasing number of applications in energy networks? At the larger end of the supply chain, advances in machine learning are bringing improved forecasting of renewables production for the benefit of system operators and generators. This is well illustrated by a 2017 Network Innovation Allowance project between National Grid ESO and The Alan Turing Institute, which contributed to a multi- model ensemble forecast more accurate than previous day-ahead forecasts for solar power. Customer-facing businesses are also increasingly able to deploy AI in order to offer more personalised services. Advances in natural language processing and question answering systems, for example, are powering the increased deployment of AI-enabled virtual customer service assistants. Arguably one of the best opportunities for AI in the future energy sector lies with customers, however, amid the current trend towards democratisation of the grid. At the domestic level, new energy technologies including rooftop photovoltaics, battery storage, electrified transport and smart thermostats are being adopted. In parallel the home is being increasingly digitised, for example with the rapid adoption of smart assistants capable of integrating sensors and home automation. The interactions between these technologies and AI capabilities such as planning, scheduling and optimisation open up the prospect of entirely new models for energy and related services. Widely anticipated possibilities include peer-to-peer energy trading and the ability to engage with a highly differentiated range of customer tariffs. Under such arrangements, customers may also choose to pool resources to enable the sharing of any associated risks. While the deployment of AI to properly harness such opportunities may still seem out of reach, it remains modest when compared to the long- term goals of the technology. Beyond energy networks there have been other impressive recent successes driven by simultaneous advances in computer hardware, availability of data and theoretical understanding of AI. In 2016, for example, a collaboration began between Google DeepMind and Moorfields Eye Hospital on the interpretation of three- dimensional eye scans. These scans had required time-consuming, detailed study from a highly trained retina specialist. In the project, a team of opthalmologists and optometrists helped train deep neural networks to recognise 50 common eye problems. While the system's impressive accuracy is on a par with world- leading consultant ophthalmologists, the transformative aspect is the fact that these otherwise time-consuming diagnoses can now be produced within seconds without the need for human input. Of course, the advance of AI comes with challenges. Issues of privacy and ethics must be properly handled, for example where customer data is used to train algorithms. Reliability and trust will be key, and human verification is still used to verify AI medical diagnoses. Work is also ongoing to improve the interpretability of AI algorithms, so that both their strengths and potential limitations may be better understood. Nevertheless it is clear that the deployment of AI offers the prospect of improved speed, accuracy and scalability in a variety of energy network applications. tion whereby artificial intel- ligence was part of a system able to identify thousands of previously un-located electric vehicles and solar panels. The network equipment is failing," he says. "Computers can cross check a lot of data from a lot of sources and determine trends that humans would not be able to see." Network operators need help dealing with challenges both in demand for electricity and gen- eration of it, Brazier explains. "We are starting to see new types of demand such as electric vehicles and potentially heat pumps. These are significant demands that the networks were not designed for when they were put together in some cases more than a century ago. The chal- lenge is ensuring that we don't have to build a much bigger network to enable people to use these devices; the cost would be unacceptable." NETWORK / 28 / OCTOBER 2019