AI-Powered Solution for Utilities

This article, written by Sébastien Cajot, was originally published in Bulletin.ch magazine in May 2024. It was then translated from French to English by Urbio.

Decarbonizing the Built Environment Through Generative Design

With the majority of buildings still heated by fossil fuels and a renovation rate of just 1% per year, is it still realistic to achieve climate targets? Fully digitalized planning powered by AI could address the workforce shortage by prioritizing high-impact projects.

Utilities play a key role in decarbonizing buildings, whether through deploying district heating networks and solar panels, or by supporting cities in developing territorial energy plans. However, with two-thirds of buildings still heated by fossil fuels, a renovation rate of only 1% per year, and a shortage of qualified labor, a new challenge emerges: how can we achieve climate targets at the pace and scale required? Today, innovative digital solutions enable organizations to "do more with less," making current teams more productive and allowing limited resources—financial, temporal, and material—to be concentrated on projects with the greatest impact.

Urbio, a Swiss software development company founded in 2020 as a spin-off from EPFL Valais Wallis, has developed a tool that innovates on three fronts, combining reliable data, interactive maps, and a "generative design" module in a web-based platform. This latter module, currently patent-pending, enables users to intuitively generate georeferenced energy scenarios, much like shaping textual content today with tools such as ChatGPT.

A Complex Energy System

The mass deployment of heat pumps, district heating networks, solar panels, charging stations, and other decentralized systems is making the task of energy planners increasingly complex. In the past, energy flows were relatively simple: they originated from a few centralized sources and were delivered to relatively predictable consumers. Today, roles and boundaries are blurring, particularly due to our dependence on intermittent resources and the emergence of "prosumers," who consume the electricity they produce in a decentralized manner.

Moreover, fragmented or absent data sources create uncertainty when it comes to investing in capital-intensive projects. With trillions of dollars invested globally in building decarbonization and a shrinking workforce, effective upstream planning is vital to ensure that local stakeholders prioritize their efforts on the most cost-effective projects with the greatest impact on CO2 reduction and renewable energy utilization goals.

Currently, we lack the human resources needed to implement these plans at the pace and scale required to align with Paris Agreement targets. Traditional methods are time-consuming and limit the ability to explore possibilities systematically and dynamically.

Furthermore, data used to inform investment decisions must be structured and easily accessible to enable tracking and updating of infrastructure project business plans in a dynamic context where parameters change almost daily.

Data, Mapping, and AI

Behind the scenes at the Energypolis innovation park in Valais, Switzerland, the startup Urbio has developed a web-based platform bringing together three innovative and complementary modules (Figure 1). The objective: enable utilities and engineering firms to identify the most attractive opportunities ten times faster than before. This platform allows users to identify the best opportunities nationwide, then size energy systems and their techno-economic parameters, all built on a turnkey, customizable database tailored to user preferences.

Figure 1: The three pillars of Urbio's 3-in-1 software.

Each module serves a specific function. The "data factory" centralizes and combines over 80 different data sources, including open data sources. Machine learning model attributes and user databases are added to this foundation. While historically, manual data collection and processing could represent up to half of an energy planning budget, this solution can reduce costs to less than 1% of the total budget.

The "digital twin" acts as a Google Maps dedicated to energy data. Instead of searching for gourmet restaurants or pedestrian routes as on Google's platform, users connect to the digital twin to find information such as a building's heat demand or its roof covering. While this type of exploration can be done today on traditional geographic information systems—requiring specialized expertise—Urbio democratizes these tools so that entire technical and commercial teams can benefit from the added value of visual and interactive maps without advanced training. Another advantage: this digital twin enables exploration, filtering, and aggregation of millions of buildings simultaneously, on the fly.

Finally, "generative design," powered by artificial intelligence (AI), automates the creation of georeferenced scenarios based on user objectives and constraints. Leveraging a complete and transparent dataset, the model provides accurate results in minutes. Where manual scenario generation can take more than 40 hours—between data preparation and problem modeling—this module produces similar results in 5 minutes.

AI as a Work Companion

Beyond the underlying algorithms, Urbio also innovates in how users interact with the platform to produce the desired results. Although close in name to "generative AI," the solution's generative design differs from the Large Language Models (LLMs) employed by tools like ChatGPT. While these tools excel in verbal interactions with users for generating creative textual or visual content, results remain largely inexplicable and limited when working with numbers or spatial information. Conversely, Urbio is patenting a solution that produces optimized, explainable, georeferenced net-zero scenarios while enabling intuitive user interaction directly on a map. Thus, rather than relying on the machine to produce a "final solution"—which rarely matches real-world complexity—users shape their plan iteratively, incorporating their expertise and contextual knowledge, while algorithms recalculate indicators on the fly, optimizing various objectives such as CO2 emissions, investment costs, or other targets. The user remains in control while their digital "companion" supports their actions in the background.

By leveraging generative design, energy stakeholders can screen larger territories much faster than before. Results are improved because different options can be compared at lower cost, increasing return on investment.

What Are the Benefits for Utilities?

District heating networks are a hot topic for many Swiss utilities, including Romande Energie, Groupe E, and Oiken. In German-speaking Switzerland, the utility CKW wanted to know which areas should be prioritized next for district heating network deployment in its service territory. Facing the need to act quickly in a dynamic environment, the company sought to improve its process.

Figure 2: CKW used the complete 3-in-1 platform suite to streamline and improve district heating network feasibility studies.

In the past, CKW's assessments involved significant manual effort: collecting building attributes, mapping baseline data in desktop GIS tools, then adjusting thermal demand models in ad hoc spreadsheets. Processes suffered from incomplete data and repetitive work. With the presented solution, CKW was able to automate the entire process, saving valuable working hours and enabling the team to focus on exploring scenarios with the support of turnkey maps and key indicators (Figure 2).

Figure 3: Three alternative district heating network scenarios created with AI (indicative data for confidentiality reasons).

The main advantage of this fully digitalized approach lies in its ability to compare variants addressing different objectives and retain only those corresponding to targeted strategic goals. Figure 3 illustrates how three indicative scenarios—generated on the fly by AI according to user objectives—can be easily compared. Scenario 1 presents an ambitious solution in which all buildings on the site are connected to the network. Greenhouse gas emission reductions are greatest here, but so are investment costs. Conversely, Scenario 3 only aims to connect one building type—in this case, oil-heated buildings. In this scenario, nearly 84% of emissions can be saved by connecting only 14% of buildings for about one-tenth of the investment costs, representing a strategic solution if that is the targeted objective. Ultimately, the decision belongs to users, who can explore more than a hundred criteria available in the platform and thus optimize investments.

---

Want to add AI to your district heating planning?

👉 Try Urbio for free or talk to our team about getting started.