Micro-grid Development Using Artificial Neural Network for Renewable Energy Forecast and System Control

Abstract

This article presents an approach to develop an electrical micro-grid with more than one renewable energy source. The micro-grid is considered as a discrete power system composed of distributed energy sources and different kind of system loads. The loads of the micro-grid can operate in alongside with the grid or can switch to other power system to supply continually the electrical energy, otherwise the micro-grid power system can switch to an accumulating reserve, such as battery banks, fuel cells or other allocated reserve. Operation of the micro-grid relies on demand, storage, and energy generation management. In this case the energy supply comes from the renewable energy system or from the accumulated energy reserves. The proposed development comprehends the use of a series of data, from the monitoring and data acquisition unit, process the data by constructing and train the neural network with this data. After the validation of neural network error rate, we use the network function to estimate a short term time series prediction. The data input is a daily solar energy value, which was measured a period of time. From this data we try to forecast the energy production value for the next day. The predicted value may serve to control the renewable energy based power system, in function of energy demand. In the studied system we have a combined hybrid power generation unit, a solar photovoltaic and a micro hydro, where the main sub system is the solar photovoltaic and the micro hydro is serving as an accumulative reserve, partially controlled by the estimated predicted data.