Overview / Usage

Fixed Point solar-plates can be automated to improve the efficiency and increase the overall power throughput. This can have positive-potential impact towards environment, and more green energy can be generated. We can have a series of solar plates/panels which can be driven all together at different orientations by a RL-agent trained using Deep Reinforcement Learning algorithms with small and efficient neural net.

Methodology / Approach

Policy gradient (Deep) method will be used to train the RL-agent. 3D-orientation, sun's position, time-stamp etc. can be taken as states. This setup is environment, where the RL-agent will take actions such as rotating different solar plate, changing their x, y, z coordinates. Power throughput can be taken as rewards and complete Markov-Decision model tuple. Once DRL network is trained, it will send action signals to various solar plates to position, rotate etc to achieve maximum efficiency.

Technologies Used

1. Anaconda Distribution
2. Intel Xeon processors
3. Intel Movidius
4. An ECU
5. Solar Plates/Panels/arrays