Forecasting Solar and Wind data using Dynamic Neural Network Architectures for a Micro-Grid ensemble

The use of renewable sources of energy is encouraged due to fast reduction in conventional non-renewable energy sources. However, finding new installation sites for power generation and transmission has become increasingly difficult. The need for more flexibility in electric systems has led to a new concept in power generation - Micro-Grid. A Micro-Grid is defined as an integrated power delivery system consisting of interconnected loads, storages facilities and distributed generation mainly composed of renewable energy sources. This paper presents a dynamic model of Micro-Grid ensemble simulated in MATLAB Simulink and the applicability of Dynamic Neural Network Architectures for forecasting Solar and Wind generation data. In total, three architectures have been proposed, namely - Focused Time Delay Neural Networks, Distributed Time Delay Neural Network and Nonlinear Auto Regressive Neural Network. The experimental results show that all the proposed networks achieved an acceptable forecasting accuracy. In terms of comparison, highest forecasting accuracy was achieved by Distributed Time Delay Neural Network.