CIPA: Predicting the Cost of Connecting to the Grid with ML
Introduction
Grid interconnection is the invisible bottleneck of the energy transition. A solar or wind farm can be designed, permitted, and financed in relative good order, only to stall—sometimes fatally—when its developers discover that the cost and timeline for connecting to the nearest available substation are far higher than anticipated. Across the United Kingdom, and particularly in Scotland, where some of the richest renewable energy resources sit in the most grid-constrained geographies, interconnection uncertainty has been one of the leading causes of project failure for over a decade. The UK Parliament's Scottish Affairs Committee has described grid connection charges as actively "hampering" the transmission of Scottish renewables, and in some of the most resource-rich regions like the Highlands and Islands, the wait for a bankable grid connection date can stretch across multiple years.
Against this backdrop, a small team at the University of Edinburgh set out in 2014 to build something deceptively simple: an algorithm that could rapidly estimate the cost of connecting a hypothetical power plant to the grid, using open data, satellite imagery, and graph search techniques. They called it CIPA—the Cost of Interconnection Prediction Algorithm. What followed was a five-year R&D journey that would illuminate both the extraordinary potential and the sobering limitations of applying digital innovation to deeply analogue infrastructure problems. CIPA would ultimately attract £500,000 in Innovate UK Smart Grant funding in 2020 and be incorporated into a venture-backed renewable energy software platform. But the path from doctoral research to commercial product was neither linear nor assured, and the lessons it offers are relevant to anyone attempting to bridge the gap between university research and market-ready climate technology.
The Origin: Where Sustainability Met Software
The story of CIPA begins not in a lab, but in the informal innovation ecosystem that had grown up around the University of Edinburgh in the early 2010s. The project's lead founder, Phillip Bruner, was a renewable energy project planner who had already co-founded two earlier sustainability ventures—a governance-focused research forum, the SGRF and SCENE, a social enterprise providing consultancy and data services to community energy projects in Scotland. Both had yielded valuable experience but limited commercial traction. Through SCENE, however, the founder had gained hands-on knowledge of the geospatial data tools and open-source energy datasets that were beginning to emerge from academic and government-funded research initiatives across Europe. He had also encountered, viscerally, the frustration of planning wind farm and small-hydro projects in the Scottish Highlands only to be blindsided by grid connection costs and timelines that no amount of site-level analysis had prepared him for.
A pivotal turn came through Edinburgh's TechMeetup, a monthly gathering of developers, researchers, and entrepreneurs hosted at Appleton Tower, home to the University of Edinburgh's School of Informatics. TechMeetup was co-sponsored by Skyscanner—one of Edinburgh's most celebrated startup success stories—and Informatics Ventures, and had developed a reputation as a place where technical and non-technical founders could find each other. Inspired by the format but noticing the conspicuous absence of sustainability-focused participants, Bruner launched GreenTech Meetup in late 2014, hosted at the Edinburgh Center for Carbon Innovation. Over approximately twelve months, the group convened six events averaging around forty attendees each, drawing from both the sustainability and technology communities. It was at one of these events that Bruner met Chris Nater, a University of Edinburgh computer science graduate with experience at CERN, the European Space Agency, and several UK tech startups. Over the course of several years with support from the ECCI, Nater and Bruner collaborated on building analytical dashboards that aimed to standardize solar and wind due diligence by attempting to align investor and developer checklists. While these early iterations culminated in a partnership with the Open Data Institute, it would take several years and the departure of Nater for Bruner and a junior team from Imperial College London to find a more specific use case for open energy data with a more clearly defined value proposition.
The Problem: Grid Opacity and the Cost of Not Knowing
Bruner's investigation eventually converged on a single, mostly technical challenge that stood out among the many barriers to renewable energy deployment: the difficulty of predicting grid interconnection costs.
The problem was structural. In Scotland, District Network Operators possessed detailed capacity and infrastructure data as a function of their core business operations, but this data was largely opaque to the developers, investors, and consulting engineers who needed it most. RE firms attempting to site new power plants had to rely on an ad hoc patchwork of information sources: limited hosting capacity maps published by some DNOs, out-of-date satellite imagery from Google Maps, and expensive proprietary data from vendors like Airbus and Maxar. The manual process of designing an interconnection pathway—identifying the nearest viable substation, routing a cable, and estimating costs—typically required teams of six to eight people working over weeks or months for a single project, at an estimated cost that one senior industry contact put at roughly £720,000 per year per firm.
Worse, the data that was openly available was what energy researchers had come to call "tedious data": mislabelled, multilingual, inconsistently formatted, frequently incorrect, and almost never machine-readable. Power plant names might be listed under parent companies, subsidiaries, or joint venture entities interchangeably. Coordinates might be wrong. Entries might be duplicated, overloaded, or simply missing. For a team trying to build software that could automate any part of this process, the data quality problem was not a secondary obstacle—it was the primary one.
Through a combination of online research, academic literature review, and confidential conversations with former industry partners, Bruner mapped out the commercial implications. RE firms were spending enormous sums on manual data collection and analysis that was, by their own practitioners' admission, often based on regional average costs rather than site-specific modeling. Projects that passed feasibility review were being abandoned after grid connection realities emerged. Investors were citing grid uncertainty as a top-tier risk factor. The wasted capital and labor were not marginal—they were a structural drag on the pace of the UK's energy transition.
The Technical Approach: Open Data, Image Recognition, and Least-Cost Paths
What Bruner and team proposed was a web-based application that could semi-automate the calculation of grid interconnection costs and optimize route selection for commercial-scale solar, onshore wind, and hydroelectric power plants. The approach would combine several technical components into a single user-facing tool.
First, the team would build a proprietary database of electric grid infrastructure derived from openly available sources—principally Open Street Map—supplemented by open power plant and substation data from initiatives like Open Power System Data, Enipedia, and OpenEI. To address the data quality problem, they would develop optical image recognition techniques, adapted from open-source methods originally designed to identify wind turbine coordinates from satellite imagery, to validate and correct grid-level data including substations, transmission towers, and overhead lines.
Second, they would implement graph search algorithms—specifically Dijkstra's algorithm applied to a grid of terrain cells—to generate least-cost cable routing paths between a proposed power plant site and its nearest viable grid connection point. Each cell in the grid would be assigned a traversal cost based on terrain type, land use, and other geospatial parameters, allowing the algorithm to compute an optimized route that accounted for real-world physical constraints.
Third, the cost calculator itself would be configurable: users could input their own parameters for cable specifications, utility pole spacing, excavation costs, and terrain multipliers, reflecting the reality that supplier costs varied significantly across projects. The system would take three inputs—generation capacity, transmission voltage, and project coordinates—and output an estimated total cost of interconnection, including cable costs, laying costs, and land use charges.
The collaboration with the University of Edinburgh's School of Engineering, led by Professor Daniel Friedrich, provided essential expertise in grid simulation and data engineering. The Data Lab, Scotland's innovation center for data science, contributed additional analytical support. Together, the team spent approximately twenty-four months developing and testing a working prototype.
The Innovate UK Investment
In late 2020, the project secured a £500,000 Smart Grant from Innovate UK—a competitive award reserved for commercially viable, innovative or disruptive ideas. The grant funded a nineteen-month collaboration between the the commercial venture ENIAN (which had by then incorporated the CIPA work), the University of Edinburgh School of Engineering, and The Data Lab, running from December 2020 through May 2022. The stated aim was to digitize, automate, and enhance the way project planners estimate grid connection costs, using machine learning and open-access data.
The timing was significant. Grid interconnection had been identified as a critical constraint on the UK's net-zero ambitions by bodies ranging from Ofgem to the Scottish Affairs Committee. Research from Lawrence Berkeley National Laboratory in the United States would later document starkly similar dynamics across American electricity markets: interconnection queues more than double the total installed capacity of the US power plant fleet, wait times up seventy percent over the preceding decade, and withdrawal rates of eighty percent—driven in significant part by the unpredictability of interconnection costs.
CIPA was, in other words, addressing a problem that was not merely Scottish, not merely British, but global in scope—a structural feature of how electricity grids had been designed, governed, and made legible (or not) to the market participants trying to build the energy system of the future.
The Prototype: What It Could Do
The CIPA prototype web application enabled users to input a proposed power plant's coordinates, capacity, and voltage, and receive an estimated cost of grid interconnection. The system automatically detected the closest substation, visualized start and end points on a map, generated a heatmap of terrain costs, and computed a least-cost overhead line routing path. Users could alter terrain multipliers to generate and compare alternative routing scenarios.
The prototype's outputs closely resembled existing overhead lines visible in Open Street Map data—a meaningful validation of the underlying approach. Its relatively simple configuration and intuitive interface allowed users to assess the causality of the algorithm's decisions and to manipulate paths to avoid obstacles or high-cost areas. For an industry accustomed to weeks of manual due diligence to produce a single interconnection cost estimate, the prospect of generating comparable outputs in hours represented a potential savings the founders estimated at twenty weeks and approximately £500,000 per year per company.
Epilogue
CIPA was ultimately integrated into ENIAN's Project Manager software as a core component of its siting intelligence feature. The grid interconnection problem that CIPA sought to address has, if anything, intensified since the project's inception. Scotland's grid infrastructure, designed for fossil fuels nearly a century ago, continues to constrain the deployment of renewable generation at precisely the moment when the country's net-zero targets demand acceleration. The tools and techniques that CIPA pioneered—applying graph search algorithms and open geospatial data to automate interconnection cost estimation—remain technically sound. Whether they can be made commercially viable within the institutional and regulatory constraints of the energy sector is a question that remains open for the next generation of climate-tech entrepreneurs.