ADIOS - I-NERGY Demo Application
Demo application deployed on our server
Main Characteristic
The demo application is developed using Python programming language on backend and React/Redux UI. It enables users to import training and full datasets. Data is parsed and stored into TimescaleDB using a multiprocess importer. Data is then processed using machine learning to detect anomalies by training the machine learning model using expert knowledge labelled data set. Finally, data is visualised using Grafana open source analytics and interactive visualisation web application.
Research areas
Collaborative AI
Technical Categories
AI services
Business Categories
Energy
Last updated
13.10.2022 - 20:04
Detailed Description
The application provides following services which can all be invoked using UI and are implemented as rest API
- Upload tagged template - Used for providing the system with a template which contains parsing rules for tagged dataset to be imported into database
- Upload tagged dataset - Enables user to upload initial training dataset which was labelled by domain expert
- Extend tagged dataset - Service which extends labelling of the training dataset by labelling all the records using smallest distance algorithm
- Benchmark training - Enables user to benchmark training algorithm before running full training process
- Train model - This service starts ML training based on extended training dataset and outputs ML models
- Upload full template - Used for providing the system with a template which contains parsing rules for full dataset to be imported into database
- Upload full dataset - Enables user to upload full dataset which will be labelled by ML algorithm using models generated in training phase
- Benchmark applying model - User can benchmark applying of ML model before the full process is executed on potentially very big dataset
- Apply ML model - Applies models generated in training phase to full dataset
- Analytics - provides visualisation of the data imported as well as results of ML labelling
Trustworthy AI
There are no particular risks associated with the PoC we will develop at this stage, however we think that TSOs that will rely on the AI models for anomaly detection need to avoid human deskilling and total reliance on the system. These implications have been further analysed during the project. Proposed mitigation actions have been put in place to reduce ethical risks.
Trustworthy AI Assessment
European Commission ALTAI