Overview
Since 1984, Klimamichaniki has been a dynamic force in Greece's HVAC and energy sectors, consistently leading the way with innovative approaches. The company's most recent initiative focuses on harnessing data and employing data-driven methodologies. This strategy aims to craft tailored solutions, provide sophisticated insights to clients, and draw in new customers, reinforcing Klimamichaniki's commitment to staying at the forefront of the industry.
The Scientific Challenge
Development of machine learning models is not an easy task for non-experts. Klimamichaniki has vast experience in the energy and HVAC domain, but is new to implementing machine learning and artificial intelligence algorithms. Collecting data from an installation is the first of many steps that takes to create an AI model, and Klimamichaniki is looking for partners and consortia with expertise on the matter, in order to be able to develop robust solutions. Additionally, given that the scope of the pilot is to create a predictive maintenance model, a significant challenge will be the absence of failure events for the cooling units within the duration of data collection. This challenge will be overcome by utilising other datasets, in order to understand the mechanics behind predictive maintenance, and possibly utilise this information for transfer learning purposes.
Who benefits and how?
- OpenAIRE EXPLORE: Academic research - OpenAIRE NEXUS
- EGI Notebook: Develop code - EGI Foundation (EGI-ACE)
- DEEP training facility: Train models
- ARGOS: Create data management plan - OpenAIRE NEXUS
- B2SHARE: Networking and dissemination of results - EUDAT (DICE Project)
PreMaCOOL has a sequential implementation plan, common in most research works, study the bibliography and market, try the solution in mind, adhere to any obligations and restrictions and finally share the results. Therefore the logical way to implement EOSC services was to utilize OpenAIRE EXPLORE for academic research, and use search engines to search for similar existing solutions. Then, access to tools and resources were required to develop code, implement ideas, train models, and test their results., therefore EGI Notebook and Deep Training facility were a perfect fit. In fact, most of the PreMaCOOL effort was spent using these two services. Then, PreMaCOOL utilized ARGOS to create a data management plan, which is necessary since the data that were used were not public therefore a data management plan might come useful in the future. Last but not least, B2SHARE was selected to disseminate the results of PreMaCOOL.
Technical Implementation
PreMaCOOL is based on a linear workflow, based on sequential and parallel activities. On the technical aspect, PreMaCOOL will be based on the following activities:
Preprocessing:
- Setting a communication mechanism in order to retrieve data
- Align all dataset variables based on rounded timestamps
- Investigate for anomalies and gaps, and clean the dataset
- Perform normalisation when necessary
- Restructure data for proper training
Training:
- Investigate algorithms
- Train model
- Fine tune selected model
- Select performance metrics
In parallel, PreMaCOOL’s dissemination activities will be taking place throughout the duration of the pilot. Social media posts, website announcements and other forms of communication will be part of the pilot in order to disseminate the company’s research activities. Direct communication, still a prominent means within Klimamichaniki’s network, will also take place, in order to show the company’s new profile, as well as to understand the clients’ interest and intent, to utilise such a service.
By the end of the pilot, Klimamichaniki will have valuable results to demonstrate, and a clear path on how to proceed with the commercialization of such a service.