The project is a Master of Science project in partial fulfilment of requirements for MSc. The objective of the project is to explore different kinds of energy management solution in grid-connected microgrid in order to maximize the return.

The project is carried out in three phase:
1. Phase 1 - comparison of different forecasting techniques
In any energy management solution, an accurate forecasting solution enhances the quality of the energy management control. In this project, 4 forecasting neural network architechture used in sequential data are compared in term of the forecasting accuracy, namely 1) normal LSTM, 2) sequence to sequence, 3) sequence to sequnce with attention and 4) the cutting edge transformer. The best architecture is deployed in the phase 2 for prediction of PV (solar), energy load and real time price

2. Phase 2 - comparison of optimizaton techniques
In this phase, we compare the performance of two algorithm - 1) the reinforcement learning (Q-learning) without any future knowledge, 2) the model predictive control with the forecasted data

3. Phase 3 - user graphical interface

** Note that the original script is coded in google colab, thus changes to all the model weights or file directory are needed in the event of reuse of code