Savonia Article: Smart heating systems in a climate-friendly way by utilizing artificial intelligence (ÄLLITÄ)
This work is licensed under CC BY-SA 4.0
The goal of Finland’s national energy and climate strategy until 2030 is to move the country towards climate neutrality by significantly increasing the use of renewable energy. The goal is to increase the share of renewable energy to at least 51 percent of the total energy consumption and to reach 30 percent of the final energy consumption of road traffic. The Just Transition Fund, based in Pohjois Savo, plays a key role in this change by connecting provincial leaders with added value in key industries, focusing on issues such as digital innovation, skills development, and workforce development. This project is specifically aimed at companies in the energy sector and promotes the use of energy and environmental data to integrate digitization and artificial intelligence (AI) into their own operations and in line with the industry 4.0 paradigm to increase productivity and sustainability.
The project has three main work packages: project management, networks, and communication (TP1); data and artificial intelligence in smart energy solutions (TP2); and artificial intelligence as a replacement for routine tasks (TP3). The main focus is to use IoT solutions and artificial intelligence to improve the energy efficiency and automation of buildings, while developing predictive models for the performance of local renewable energy such as solar energy. In addition, artificial intelligence is used to innovate business processes in the energy sector, and applications are expanding to other industries as well. Examples include using AI to streamline routine tasks such as structuring abstract writing ideas, creating new product concepts, and generating software code. The aim of the project is to use these initiatives to raise awareness of artificial intelligence in Northern Savo and improve the competitiveness of companies, strengthen the digital competence of the region and support DigiCenterNS as an artificial intelligence and IoT competence center.
Data and AI in smart energy solutions
Data is the backbone in the field of AI and decision making. Our first pilot regarding electricity efficiency is Savonia-ammattikorkeakoulu Oy, Varkaus. For ÄLLITÄ project, we need as much data as we can get from diverse sources. Currently, we have electricity consumption (kWh) data from January 2023 – April 2024 on an hourly basis. This dataset had only two features i.e., datetime and consumption (kWh). We performed exploratory data analysis to understand electricity consumption patterns. Figure 1 shows the electricity consumption from Jan 2023 – Apr 2024 on an hourly basis.
Figure 1: Electricity consumption (kWh) over time.
We performed feature engineering on this dataset and created suitable features to understand the consumption patterns over time. Figure 2 displays the average electricity consumption by month, day of the week, and hour of the day, shown from left to right. Average electricity consumption is lower than 25 kWh from May-Aug while it is between 25-30 kWh during Sep, Oct, Nov, Jan, Mar, and Apr. November and February are the peak months where average electricity consumption goes above 35 kWh . We can hypothesize that this peak is due to the extreme winter for the season 2023-2024. Electricity consumption is lower during Saturday and Sunday. Electricity consumption is high during the time from 03:00 to 17:00 of the day due to the scheduled set in HVAC system of the building and it stays on peak from 08:00 – 15:00. We can hypothesize that this increase in electricity consumption is due to the working hours of the university.
Figure 2: Average electricity consumption by month, day of week and hour of day.
We also extracted features for weekdays and holidays from the dataset. Holidays are the official holidays of Finland and weekends consist of Saturdays and Sundays. It is evident from Figure 3 that electricity consumption is lower during weekends and holidays. During weekends, average electricity consumption is almost 25 kWh while during holidays it is slightly above 25 kWh.
Figure 3: Electricity consumption w.r.t working/non-working day.
We also have a dataset of weather-related features from the Finnish Meteorological Institute (FMI) on an hourly basis. We also have heating, ventilation, and air conditioning (HVAC) data as well as data from the district heating company regarding indoor conditions of the building. Our next step will be to extract suitable features from all the datasets and experiment with different data-driven AI models to predict future electricity consumption. We will also build data-driven AI models to predict solar power production. We will use our NVIDIA DGX A100 AI server for model training, testing and deployment.
Code Generation through AI
As mentioned before, the third work package (TP3) is concerned with code generation for simple tasks. The idea behind this is to create a tool that helps companies complete menial tasks in a faster manner. This package highlights efficiency in the workplace as an important goal. By using Artificial Intelligence to generate small code snippets or debug code, companies and their employees can save time on otherwise repetitive tasks. Another sector of interest for this package will be Web Development and exploring the limitations of code generation for creating user interfaces. On top of the Software Development approach, the tool we are aiming to develop will help with the creation of presentations, product concepts, articles, and other written content. Our approach will utilize the capabilities of open-source AI models, which will be running on the NVIDIA AI server. Due to the high amount of resources that the AI server possesses, we have a lot of freedom as we can host almost any AI model, no matter the requirements.
Authors:
Shahbaz Baig, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, shahbaz.baig@savonia.fi
Premton Canamusa, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, premton.canamusa@savonia.fi
Mika Leskinen, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, mika.leskinen@savonia.fi
Aki Happonen, project coordinator, DigiCenter, Savonia-ammattikorkeakoulu, aki.happonen@savonia.fi