Overview / Usage

The project focuses on the prediction and optimisation of the charging and connection process of the batteries of a solar power installation (small installations). This process takes into account not only environmental factors but also the configuration of market prices for consuming/sending energy to the grid.

​

Methodology / Approach

We will use data from Local Energy Markets and Copernicus weather data to train a bayesian neuronal network. This NN will be used jointly with Markov models (energy market) to get a mix of experts in order to get an optimized prediction system. Thi system will be deployed to Ai devices (low cost as possible) which will be used in local Photovoltaic installations.

Technologies Used

GPU/FGPA systems for training (Bayesian NN)

Intel IoT devices with IoT capabilities (or NVIDIA)