A Critical Analysis for Microgrid Formation based on Hourly Load Growth and Available Renewable Energy

Abstract

In this 21st century, Microgrid is becoming prevalent in distribution System because of a number of benefits that it has. A lot of work is going on to make the integration of microgrid in the present system easier. Our conventional power generation concept is the generation following the load profile and the distribution system is passive in nature. But on integration of the microgrid concept in the distribution system, it becomes an active system. Due to the incorporation of Distributed Generations (DGs), the idea becomes load following the source. It is therefore essential to have the information of availability of the renewable resources and installed generation capacities in advance. The advantage of microgrid lies in the fact that it can operate both in utility grid connected mode and also in autonomous mode. However, the intermittent nature of the renewable resources demands advanced knowledge of their availability for power management within and outside the microgrid. This requires proper prediction of resources a day ahead of actual operation. Wind and solar energy have the highest eligibility as renewable energy resources. Many forecasting techniques are developed and utilized for the prediction of the availability of these energies. In this field of forecasting, Artificial Neural Network (ANN) has proved itself to be very competitive. Long Short Term Memory (LSTM) network is the recent improvement of ANN which has been applied in this work for forecasting both solar and wind energy. This is compared with the forecasting using multiple feed forward neural networks. It is seen that LSTM performs better for day ahead forecasting.