Notice

AI for Decarbonisation Innovation Programme: Stream 3 successful projects

Updated 19 March 2024

Microgeneration energy export optimiser

Lead applicant: Open Power
Amount awarded: £313,700

Description: We are creating software which allows sites that generate their own electricity to participate in the UK’s local and central energy markets. These sites could be a residential home, school or local business who have solar or micro-wind installations.

By solving the technical challenges of selling their export electricity across various energy markets, we can maximise the rates they receive for the energy they export to the grid. This can considerably reduce the payback period of micro-renewable installations, making them a better financial decision, and thus speeding up their adoption. This accelerated adoption will help towards this country’s net zero targets.

OnGen expert AI pro

Lead applicant: OnGen Ltd
Amount awarded: £326,371.08

Description: OnGen delivers proven, award-winning digital solutions to facilitate the decarbonisation of consumers’ energy supply.

This project builds on the successful delivery of an Innovate UK supported project that introduced AI into the process of optimising behind-the-meter renewable energy systems, demonstrating to the energy consumer how to generate the best return on investment and modelling how third-party finance can be utilised to fund the capital expenditure required.

AI as transport decarbonisation enabler: removing commercial fleet electrification barriers through energy use optimisation

Lead applicant: Flexible Power Systems Ltd
Amount awarded: £209,360.35

Description: FPS Operate is an automated platform that dynamically schedules electric vehicle fleet operations and their charging in real time. The platform has been in operation with a number of users since 2021. Our software considers factors such as:

• which vehicles need charging and the state of charge of their batteries
• forecast operational requirement
• available charge points
• local grid conditions in terms of agreed site load and forecast energy consumption
• time of use pricing

We enable smart allocation of vehicles to routes, based on vehicle route plans created by existing routing software.

This project will enhance the current state of art of integrated EV charge optimisation by deploying deep reinforcement learning techniques to better serve fleet operators and to expand the feasibility of the EV transition to more challenging sites and operations. This includes the capability to better handle:

• large multi-modal fulfilment centres with hundreds of vehicles
• smart distribution grid integration by optimising fleet charging across multiple sites connected to the same transformer, or a transformer one level up in the distribution hierarchy
• co-solving the general combined charging and vehicle-to-task matching problem taking vehicle capability and task energy needs into account

Adding task-based route allocation to the problem will allow the AI to take the decision on which energy (stored in each vehicles’ battery) will be used to execute each job, producing an optimal allocation of energy resources to the set of tasks at hand at any given moment across multiple fleets and sites. This is a critical innovation required to realise the decarbonised transport fleet system. By combining the charging problem and allocation problem the AI will be able to maximise the amount of green energy which is used to service the set of transportation tasks while ensuring a robust operation without a requirement for over specified hardware (charger rating and battery size).

This will put FPS, and the UK, on an advanced innovation trajectory, allowing transition at large scale and with complex variables, enabling further capabilities like:

• energy optimisation based on V2G (energy and ancillary market participation, distribution grid congestion relief)
• local and grid-wide renewable energy availability
• the general combined vehicle routing and charging on-the-go problem
• prioritised charging based on user preferences and task attributes like urgency

WAM (Wind farm area minimisation)

Lead applicant: EDF Energy R&D UK Centre Ltd
Amount awarded: £23,586

Description: Project WAM will make use of machine reasoning and AI technologies to bolster existing software capabilities to optimise the layout of offshore wind farms through the creation of novel functionality to minimise farm footprint, whilst also minimising the penalty the developer pays in lost energy production due to increased wake interactions.

There are numerous incentives and benefits to decreasing the area coverage of offshore wind farms. One is to make more efficient use of seabed space. As the prevalence of offshore wind energy continues to increase alongside demand for new installed capacity around the globe in line with upward-trending renewable energy targets set by governments, effective long-term marine planning to ensure space-efficient utilisation of the seabed is becoming an ever more important priority to marine authorities.

A second incentive and benefit is to increase competitiveness of bids put forward by developers in tenders. In offshore wind competitive auctions, there is evidence within mature markets that the limit to the cost reduction potential of auction mechanisms that operate on a price-only basis is fast-approaching or has already been reached. An argument can be made that this limit has even been surpassed in some markets, with the occurrence of economic phenomena such as uncapped negative bidding in the German offshore wind market. In response, auction designers are beginning to consider new ways of asking developers to compete beyond just on price. A smaller wind farm footprint is expected to afford developers a competitive advantage as more qualitative and non-price criteria are introduced into auction designs.

Developers will not however wish to reduce the footprint beyond a point where production losses become significant and outweigh any gains in bid competitiveness. Project WAM will address this trade-off through aiming to achieve the first objective of developing and testing implementations of two unique solutions to the area minimisation problem which respect user requirements on production. A second objective is to characterise the mathematical relationship between area usage and production, and to use this relationship to frame a multi-objective optimisation problem. Key deliverables include a web application intended to simplify use of and to boost accessibility to the optimisation functionality to be developed, along with a final project report. Expected benefits to be generated by the proposed innovation include:

• higher number of turbines per unit area
• increased bid competitiveness
• lower ecological impact
• increased likelihood of gaining consent
• increased sustainability of designs due to reduced cabling

Clio finance: Climate finance backbone for retail loans

Lead applicant: Clio Ventures
Amount awarded: £133,368.48

Description: Our project is designed to enable financial institutions to assist homeowners in improving energy efficiency, a critical step towards decarbonising the UK’s economy.

We are building an AI-enabled platform for lenders to create personalised banking products and streamline the validation of green projects. Our product can support a broader and faster adoption of energy efficiency measures, driving decarbonisation in hundreds of thousands of homes. This, in turn, will decrease future energy demand. Our project demonstrates how AI can be integrated into the climate finance sector to achieve a more inclusive net zero transition.

Place based AI driven clustering and demand flexibility assessment for net zero

Lead applicant: Carbon Laces Solutions Ltd, Cranfield
Amount awarded: £342,999.46

Description: This project aims to enhance the demand flexibility services (DFS) for stakeholders including the electricity system operator (ESO) of National Grid, demand flexibility providers, and consumers using Artificial Intelligence (AI), particularly machine learning (ML).

Initiated by the ESO in 2022-2023, DFS incentivises consumers to adjust their electricity usage for purposes such as peak shaving. With the ongoing transition towards net-zero, the electricity grid has been undergoing substantial updates to reduce carbon emissions, such as the penetration of uncertain renewable generation and demand surge due to electrification of transport and thermal, which poses challenges to network security.

At the heart of these developments lies the imperative to enhance the effectiveness of DFS in response to the challenges. This project will combine expertise from Carbon Laces Solutions in data engineering and demand modelling, its capacity as a DFS provider, and Cranfield University’s expertise in AI and energy systems to jointly develop an ML-driven tool for automated DFS evaluation of consumers.

The objectives include creating a prototype tool for domestic users, scaling to community-level buildings, testing the tool with consumers, and proposing a business model for future development. Key deliverables are an ML-driven DFS evaluation tool that can be applied by all UK consumers, customer experiences from using the tool, and a commercialisation business model. Expected benefits include improved efficiency and capacity for DFS provision, enhanced grid reliability, and reduced carbon emissions from peak power generation, supporting a sustainable future.

Integrated satellite and ground-based array systems with AI for enhanced solar forecasting and decarbonised grid management

Lead applicant: The University of Nottingham
Amount awarded: £263,377.73

Description: The contribution of PV to the electric grid continues to grow. In 2020, the installed capacity in the UK reached 13.4 GW, accounting for 4.1% of total electricity generation, and it is projected to increase to 40 GW by 2030.

Accelerating the adoption of solar energy presents significant challenges to the electricity transmission and distribution system. The inherent variability of solar power directly impacts the energy fed into power grids, potentially creating severe imbalances between demand and the grid’s capacity, transport, distribution, and storage. Solar power, being non-dispatchable and lacking the inertia of conventional plants it replaces, reduces system resilience to disturbances.

It’s possible to supplement PV with fast-response energy storage systems (such as batteries) to address these challenges; however, this comes with a significant increase in the PV plant cost. The requirements for fast-response energy storage can be reduced by improving the accuracy of solar power forecasting, then supporting PV with slower-response, cost-effective energy storage or alternative generation methods, thus reducing system-level costs while retaining the operational benefits of conventional generation.
 
Accurately predicting solar energy generation is a significant challenge, particularly in regions with diverse weather patterns like the UK. Solar radiation is intermittent, and the solar source at any given point on the PV array depends on the sun’s position, atmospheric aerosol levels, cloud cover, and motion.

Therefore, this proposal aims to establish a ground array network integrated with satellite images to forecast solar production with a horizon ranging from 2 minute to 72 hours. The goal is to significantly improve the accuracy of meteorological parameter predictions, thereby reducing power mismatches caused by solar forecast errors and accelerating the adoption of PV generation.

Generative AI based solution to optimize energy in non-domestic assets without a building management system

Lead applicant: Optimise-AI, Blaenau Gwent Council
Amount awarded: £125,100

Description: Buildings are complex ecosystems and have dynamic factors that impact energy performance that require a paradigm shift to gradually transform buildings from ‘passive’ to ‘actively’ operated assets that positively respond to various stimuli. This transformation needs to take into account a wide range of environmental (weather, waste disposal), behavioural (occupancy changes, socio- physiological- cultural changes), economical (building life cycle, maintenance), technical (structural, technological changes) factors as well as optimum performance considerations (energy efficiency, net zero) in order to reach energy efficiency and  for long term sustainability. Existing building management systems can only provide local adaptability by creating and managing information for a built asset (engineering aspect); however, they lack the capability to learn and adapt based on performance objectives. 

Existing building energy management solutions for medium-sized to large non-domestic buildings tend to be costly and rely on data readings provided by sensors to respond to business-as-usual scenarios, such as keeping the indoor temperature within a given range.

The digitalisation of buildings as reflected by the emergence of generative AI and digital twins present an opportunity that can effectively address the carbon agenda in a timely and cost-effective way. Current (patented) technologies developed jointly by Cardiff University and OptimiseAI Ltd already allow effective energy savings. However, 90% of non-domestic buildings are not being actively managed (non-instrumented) as they do not have a building management system (BMS) or have one that is inefficient (outdated), and most older buildings, (often the ones most in need of energy optimisation), do not have energy models or historical datasets as collecting such datasets is expensive and time-consuming.

To optimise non-instrumented buildings, we will federate AI surrogates using a semantic energy optimisation capability driven by co-learning and transferable AI adapted to socio-technical parameters of built environments. We aim to develop an energy optimisation system that provides a calibration capability based on IoT sensor readings to reduce the endemic energy performance gap, while devising pathways to achieving net zero targets.