Conserving energy using AI

Energy management using AI 

This article mainly focuses on energy management using AI. There are many methods to achieve energy management but most of them are static approaches. With a multitude of devices entering into the market the load model changes so there is a lot of power loss. This raises the need for a system that can analyse the load model dynamically in real-time and can help us regulate the supply so that we can minimize the power loss and can reduce the usage.

First, we build a neural network for finding out the optimal power ratio i.e. device/incoming. This is our target model and we need to collect a lot of data from many appliances to model it correctly. 

Then we build a new network called the model network that exactly depicts the current state of the consumption pattern. The model networks weights are very slowly modified so that it reaches the target network and this is how we train the network.  

Now the key component a discriminator evaluation policy is introduced whose sole aim is to see the optimal power ratio predicted by the network, this is useful and assures that the device or appliance caused no harm.

If the data is approved by the discriminator policy, then the network is trained. Once trained and optimal values are predicted (optimal power ratio) we then use the TRIAC based control card to regulate the AC wave at different times to get different power values. 

This is how we achieve energy optimization using AI.

Presentation Link: https://docs.google.com/presentation/d/1c5yM2pImIcD6uOiaskzgqUqm-ckXLs7eOmdHOu-cIoo/edit?usp=sharing

Project Description Link: https://docs.google.com/presentation/d/1nG6L4FdPHvI-8TQN6tlBabc5HtUxYmOYptC1lKrV-rA/edit?usp=sharing

Very Soon all the PCB files will be upload into my gitHub.

Repo Link: https://github.com/srimanthtenneti/PCB-files

*** This repo is not public as of yet due to technical reasons very soon it will be made public ***

AI's performance graphs :

The above graph is the optimized data.

Initial attempt to train the network. This still needs a lot of training.

All of the code for this project will be made public very soon. Just stay tuned and keep checking the below repository.

Repo link: https://github.com/srimanthtenneti/Energy-Management

Also don't forget to check out my articles on self-learning algorithms, computer vision and linear regression. They might help you with your projects.

Reinforcement Learning: https://medium.com/analytics-vidhya/reinforcement-learning-101-bf42523fd6ad

CNN: https://medium.com/analytics-vidhya/simple-cnn-using-pytorch-c1be80bd7511

Linear Regression: https://medium.com/analytics-vidhya/simple-linear-regression-1b15f07e555e

Thank you stay home, stay safe.