Machine Learning Enables Active Asset Management of Power Networks

Description of impact

Staff at Heriot-Watt University within the Edinburgh Research Partnership in Engineering (ERPE) led a collaborative project with Scottish Power Energy Networks (SPEN) to develop the Network Constraints Early Warning System (NCEWS), with on-board machine learning for active asset management and maintenance/reinforcement planning of the electricity distribution system. This supports the critical pathway to enable decentralised energy networks for future low carbon renewable integration. The impacts included:

(A) Integration of NCEWS within SPEN’s network management platform enabling more confident prediction of system response, penetration of renewable generation, supporting demand management and electric vehicle (EV) charging. Network design, analysis and contingency planning modelling time was reduced by 67%.

(B) Increased knowledge and closer management of the increasingly active network enabled consideration and deferral of network asset renewal and delivered operational cost savings (GBP1 billion in the SPEN distribution area alone).

(C) NCEWS was adopted in SPEN’s Digitisation Strategy, enabling the Networks to help drive carbon reduction, increase security of supply and extend asset maintenance intervals and lifetimes.

(D) SPEN were able to leverage consequent funding to upgrade an area of their network in anticipation of electric vehicle rollout. The collaboration also won several significant national awards, including the E&T 2019 Innovation of the Year Award.

Narrative

The multi-award winning NCEWS project, supported by Innovate UK through a Knowledge Transfer Partnership (KTP) designed an operational and planning decision support system (DSS) that embodied the functional, operational and compliance requirements of the SPEN distribution system. The ERPE team were able, through their underpinning research, to create advances in machine learning techniques that were integrated into SPEN’s information platform and used to support strategic network planning decisions. This has supported the critical pathway to enable decentralised energy networks for future dynamic low carbon renewable integration. The impacts include:

(A) Deployment and Integration of NCEWS within Utility Network Management Platform

The NCEWS project integrated metadata, network expertise and novel machine learning-based analysis that achieved three primary aims:

(i) It backfilled missing asset data (especially cables) in the SPEN system, by applying machine learning (ML) to detect patterns from both existing assets and smart meter (SM) data. This provided critical new network information without the disruption and expense of physical inspection.

(ii) It leveraged large-scale real-time data available from the rollout of smart meters, while preserving the privacy of individual consumers. This allowed SPEN to detect areas of the network which are prone to voltage violations and potentially poor supply quality, and to simulate the benefits of new investments, for example new electric vehicle (EV) charging stations, under large numbers of scenarios.

(iii) ERPE research was able to identify the optimal strategic SM sparse deployment locations, providing the equivalent accuracy in voltage prediction of a 100% SM installation. It accelerated network monitoring and reduced the associated costs.

The impact was increased and assured through the integration of NCEWS with the SPEN Network Analysis and View (NAVI) platform. It was rolled out and integrated as business-as-usual (BAU) across the company, to support real business and planning decisions. For example, when a network planner is unsure of the type of cable present in a particular location, he/she can inquire of the ML tool to provide a prediction that includes the degree of statistical confidence. This has allowed decisions to be taken that, hitherto, would have required physical verification (which is often expensive/unfeasible, as cables are buried underground, in complex locations).

(B) Deferral of Network Renewal and Associated Cost Savings

The use of the ML systems has also led to significant economic impact. A key benefit of predictions made by users of NCEWS is the potential to safely defer expensive network reinforcements. Recent internal figures from SPEN show that in order to accommodate the rollout of EV charging, network reinforcements could be deferred, with savings of around GBP1,000,000,000 in the SPEN distribution area alone.

NCEWS uses real smart meter data across the network and helps to pinpoint critical sub-networks, where deferrals will result in cost savings. Examples of such cost-saving decisions include smarter selection of the placement of EV charging stations, distributed storage and demand-side response, and avoiding placement in areas of likely voltage/power violations. The NCEWS predictive tool, that uses real smart meter data across the company's network and helps pinpoint critical sub-networks, is now central to achieving these deferral savings.

(C) Enabling Digitalisation Strategy and Supporting Decarbonisation

NCEWS is now being integrated into a larger national roll-out and rebranded as NAVI as part of SP Energy Networks’ ‘RIIO-RD2’ plans with OFGEM. In the SPEN Digitalisation Strategy 2019, the NCEWS project and outcomes was cited as a key project, ‘by reducing modelling design time by two thirds (67%)’, ‘automatic tracing of the network also means much larger geographical areas can be analysed…. leading to improved understanding supporting informed decision making regarding network reinforcement’.

More significant impact for NCEWS was in the standardisation of data collection across other innovation projects relating to energy networks. At the launch and introduction of the 2019 SPEN Digitisation Strategy, the CEO of SPEN stated that “Improvements in control, automation, flexibility and demand side management are helping us create a more dynamic and active network," NCEWS is assisting SPEN to support the UK's decarbonisation agenda. In some cases, where there is uncertainty whether a new investment project (e.g. building a new EV charging station) could lead to voltage/power violations, SPEN has had to err on the side of caution in the past, until the local network can be physically assessed and/or reinforced. With the aid of NCEWS, however, acting on real smart meter data, SPEN can make faster approval decisions, and faster approval/rollout of the decarbonisation investments.

(D) Consequent Funding and National Awards

The results of the NCEWS project enabled SPEN to leverage consequent funding for the PACE project, in collaboration with Transport Scotland and Scottish Government. PACE sought to accelerate and widen the installation of EV charging points across a weak area of the low voltage distribution network in Lanarkshire.

The ERPE and SPEN partnership, and the resulting development of the NCEWS, won national awards including the prestigious E&T 2019 Innovation of the Year Award.

Fiona Fulton, SPEN, Smart Grid Manager stated the significance of the project and relationship to their low-voltage (LV) network: "The NCEWS project with Heriot-Watt University was a real success, that met and exceeded our expectations. The information platform developed allows SPEN to automatically backfill missing asset data, as well as using advanced analytics to identify ‘at risk areas’ and potential voltage excursions in our LV distribution network. The project leverages the massive amounts of data made available by the smart meter rollout, and allows SPEN to be at the forefront of European innovation efforts in this key area. The ongoing Business-As-Usual rollout across SPEN will enable all parts of the business to benefit from its results, which is an outstanding result for a knowledge transfer project". The project resulted in several further awards, including:

2019 IET's Information Technology Award.

Top "outstanding" rating for a KTP project from InnovateUK,

Knowledge Transfer Partnership (KTP) "Rising Star" award for KTP Associate Maizura Mokhtar,

Overall, the project positioned SPEN at the cutting edge of electrical network modernisation and importantly serves as a keystone in their acceleration of decarbonisation and incorporation of low carbon technologies.

Impact status	Achieved
Impact date	1 Jan 2019 → 31 Dec 2020
Category of impact	Economic, Environmental
Impact level	National