In this article, we are giving same prompt to Enterprise Microsoft Copilot (Savonia’s access), ChatGPT 5.0 free edition, and Gemini 2.5 pro to extract articles and write a summary of 500 words. The prompt is “search articles from savonia.fi related to ällitä project only and write a story-based summary of 500 words”. Microsoft Copilot gave 471 words summary with a title but couldn’t extract all the articles and information in them. ChatGPT gave 740 words summary while extracting all the articles of the project and embedded links to the published articles in the text, but it didn’t follow the 500-word limit. Gemini gave 592 words summary, and it was able to extract most of the information from the published articles. Gemini’s response was more structured than ChatGPT and Microsoft Copilot as it arranged the content based on WPs.

Microsoft Copilot

ÄLLITÄ: A Journey Toward Smarter Energy and AI Innovation

In the heart of Northern Savonia, a bold vision took shape—the ÄLLITÄ Project, an EU co-financed initiative aimed at transforming energy management and digital innovation through artificial intelligence. What began as a response to Finland’s climate neutrality goals soon evolved into a multidisciplinary effort blending renewable energy, IoT, and generative AI.

The story starts with solar power. Finland’s commitment to halve peat use by 2030 and reach 51% renewable energy by 2035 set the stage for innovation. At Savonia University of Applied Sciences, a modest 2 kW solar PV system was installed on the Kuopio campus. From April to September 2024, researchers meticulously tracked its performance. The data revealed fascinating patterns—May emerged as the champion month for solar production, with daily outputs peaking at 9.19 kWh. Hourly trends showed sharp rises between 9 AM and 10 AM and declines after 4 PM, insights that would later feed predictive AI models. This was more than numbers; it was proof that even small-scale systems could contribute meaningfully to Finland’s green transition.

But energy isn’t just about production—it’s about consumption. At the Varkaus campus, the team dove into electricity usage data spanning 16 months, pairing it with weather metrics like temperature, humidity, and snow depth. The analysis uncovered intriguing correlations: colder days and deeper snow drove consumption up, while warmer temperatures eased the load. Using these insights, the team built a linear regression model to predict electricity demand. Though the initial model scored modestly (R² = 0.10), it marked the first step toward smarter, AI-driven energy solutions.

As the project progressed, the focus shifted from energy to digital empowerment. Generative AI became a star player. In one pilot, open-source models like Llama 3.3 and Bart were deployed to create a web application capable of summarizing lengthy PDFs—an invaluable tool for researchers drowning in data. But the team didn’t stop there. They pushed boundaries by integrating Microsoft’s TRELLIS, a cutting-edge 3D generation model, into the platform. Suddenly, users could turn text prompts into tangible 3D product designs, ready for download and even 3D printing. From IoT sensor casings to gaming assets, the possibilities were endless.

Meanwhile, back in the physical world, IoT devices quietly revolutionized campus operations. Custom-built systems monitored CO₂ levels to gauge occupancy, while motion sensors tracked foot traffic. Weather stations and pyranometers captured environmental data, feeding real-time insights into an on-premises Thingsboard platform. This seamless integration of hardware and AI enabled dynamic energy optimization, proving that smart buildings aren’t a distant dream—they’re here.

The ÄLLITÄ Project is more than a technical endeavor; it’s a narrative of collaboration, creativity, and commitment to sustainability. From solar panels basking under Finnish skies to AI models crafting virtual worlds, ÄLLITÄ exemplifies how technology can illuminate the path toward a greener, smarter future.

ChatGPT

In recent years, Savonia has launched an ambitious initiative — the ĂLLITÄ project — aimed at harnessing the power of artificial intelligence (AI) and data technologies to make energy use, building automation, and everyday work tasks smarter, simpler, and more sustainable. Savonia-AMK+2Savonia-AMK+2

Laying the groundwork: AI for energy-smart buildings

The journey began with the project’s core ambition: to embed IoT (Internet of Things) sensors and AI-driven analytics into building energy systems and renewable energy sources. On the campus of Savonia in Varkaus, researchers collected detailed hourly electricity consumption data from January 2023 to April 2024, while simultaneously obtaining weather data from a nearby meteorological station. Savonia-AMK+1

The combination of consumption and weather data allowed researchers to explore how environmental conditions — temperature, humidity, cloud cover, snow depth — influence electricity usage. They found that snow depth positively correlated with increased electricity consumption, while higher air temperature and dew-point temperature correlated with lower energy use. Savonia-AMK

This analysis formed the basis for a baseline predictive model: a linear regression that attempted to forecast consumption. While the model’s performance was modest (with a root mean squared error around 6.33), it set the stage for more advanced modeling — and more refined AI-driven forecasting in subsequent phases. Savonia-AMK

Going beyond electricity: Smart heating + renewables + automation

But electricity alone was just part of the picture. In the “Smart heating systems in a climate-friendly way” phase of the project, ĂLLITÄ extended its scope. The plan was to bring together IoT devices, AI, and building automation to optimize heating, ventilation, and air-conditioning (HVAC) systems — especially in the context of renewable energy like solar power. Savonia-AMK+1

At the same time, the project began exploring ways to predict renewable energy output — particularly from solar photovoltaics (PV) — and to even anticipate performance issues like 5G signal degradation (for campus networking). Savonia-AMK+1

Solar power under the microscope: reality and patterns

In one pilot, Savonia installed a 2 kW-peak solar PV system on its Kuopio campus and gathered detailed hourly production data between April and September 2024. The results revealed stark daily and monthly patterns: output peaked during daylight hours (especially between 11:00 and 15:00), and the month of May stood out as having the highest average and maximum output. Savonia-AMK

On a daily basis, the average production was about 4.9 kWh, with extremes ranging from as little as 0.11 kWh to as much as 9.19 kWh — demonstrating both the potential and variability of small-scale solar in Finland. Savonia-AMK

The project team plans to deepen this analysis by combining the solar data with weather variables and exploring AI-driven models to predict PV output — a step that could help plan energy use or heating needs more accurately. Savonia-AMK+1

Smart tools for people: automating mundane tasks

Interestingly, the ambition of ÄLLITÄ was never limited to buildings and energy systems. The project also aims to relieve people from repetitive, time-consuming tasks. One pilot already demonstrated how generative AI can be used to automatically generate software code — for example, for microcontrollers, web front-ends and back-ends — and even debug or fix layout issues (such as logo placement) on webpages. Savonia-AMK+1

An internal web-application was built: users can upload text or PDF documents, and the system — leveraging open-source models like Facebook’s BART — returns concise summaries. The tool was itself developed by AI-generated code, illustrating how far automation can go. Savonia-AMK

Smarter forecasting with time-aware data techniques

A more recent advancement within ĂLLITÄ involves refining the predictive models. Researchers found that including time-based cyclic features (like hour of day, day of week, seasonality) and lag features (past values) dramatically improved forecasts for energy consumption: in a test, the mean absolute error dropped substantially when such features were added. Savonia-AMK+1

This shift shows the project’s growing sophistication: rather than relying only on raw data, models now account for temporal patterns — a critical element especially in Nordic climates, where seasons and daylight hours swing widely.

Looking forward: a future of intelligent, sustainable operations

Together, the pieces of the ÄLLITÄ project paint a compelling vision: campuses and buildings that sense their environment, anticipate energy and heating needs, schedule operations smartly, and support sustainable energy — while also offering tools to lighten human workloads. What began as electricity-and-heat optimization has grown into a cross-disciplinary effort: renewable energy forecasting, building automation, AI-based software tools, and data-driven decision support.

Through this work, Savonia is not just testing tech — it’s building the digital and environmental infrastructure for a future where energy efficiency, sustainability, and productivity go hand in hand.

Google Gemini

The ÄLLITÄ project, officially titled “Smart heating systems in a climate-friendly way by utilizing artificial intelligence,” is an ambitious initiative spearheaded by Savonia University of Applied Sciences and co-funded by the EU. It tells a twin-story of transformation: one of optimizing building energy for a sustainable future, and another of leveraging the revolution of generative AI to eliminate routine tasks and spark innovation in business.

The first chapter of the ÄLLITÄ story unfolds across the physical landscape of Savonia’s campuses in Varkaus and Kuopio. Here, the challenge was to make public buildings smarter and greener. The solution lay in the pervasive deployment of IoT (Internet of Things) sensors. These digital sentinels were installed throughout the buildings, quietly gathering streams of time-series data—monitoring everything from real-time electricity consumption and building occupancy levels to local weather conditions. This wealth of information became the foundation for Work Package 2, “Data and AI in smart energy solutions.”

The data scientists in the project then took on the role of digital detectives, analyzing patterns in electricity use. They discovered, for instance, that consumption spiked during the university’s working hours, a pattern tied to the HVAC system’s schedule. To build highly accurate forecasting models, they dove deep into feature engineering, finding that traditional data needed advanced refinement. They incorporated time-based cyclic features (like daily or weekly patterns) and lag features (how past consumption influences the future) to significantly enhance the accuracy of their predictive algorithms. Beyond electricity, the project also monitored renewable energy generation. A 2 KW peak-capacity solar photovoltaic (PV) system at the Kuopio campus became a live laboratory. By analyzing the PV production data—observing, for example, the highest average output in May and the peak hours between 11 AM and 3 PM—they developed models to predict solar energy performance, further promoting the integration of renewables in the regional energy mix.

The second, equally compelling, chapter of the ÄLLITÄ narrative focuses on Work Package 3: “Artificial intelligence as a replacement for routine tasks.” This section explores how to equip companies with tools that streamline their day-to-day operations. Generative AI became the hero of this story, demonstrating its capability to handle everything from software development to creative ideation.

One successful pilot was the development of a functional web application for article summarization. Utilizing open-source models like Facebook’s Bart, the tool allowed users to upload PDF or text files and instantly receive concise summaries, boosting research and information-gathering efficiency. Furthermore, generative AI proved invaluable in the entire web development pipeline itself, with models like Meta’s Llama-3.3-70B-Instruct assisting in generating frontend and backend code, and even debugging complex issues. The project also showcased AI’s ability to generate code for microcontrollers, turning a time-intensive task into a swift, automated process.

Perhaps the most forward-looking pilot involved product concept generation. Using Microsoft’s TRELLIS, a massive open-source 3D generation model, users could simply provide a text prompt—such as “A fir tree” or a description of a new product—and the system would generate a corresponding 3D model. This innovative tool holds the potential to radically accelerate the initial phases of product design and concept ideation for companies.

In its entirety, the ÄLLITÄ project is more than a collection of pilots; it is a strategic effort to strengthen the digital competence of the Northern Savo region. By showcasing practical, real-world applications of AI and digitalization—from saving energy costs in public buildings to unlocking new creative potential in product development—Savonia is aiming to improve company competitiveness and solidify the region’s standing as a hub for AI and IoT expertise.

---

Authors:

Shahbaz Baig, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, shahbaz.baig@savonia.fi

Mika Leskinen, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, mika.leskinen@savonia.fi

Aki Happonen, Digital Development Manager, DigiCenter, Savonia-ammattikorkeakoulu, aki.happonen@savonia.fi

Laura Leppänen, RDI Specialist, Savonia-ammattikorkeakoulu Oy, laura.leppanen@savonia.fi

---