WSEAS Transactions on Power Systems
Print ISSN: 1790-5060, E-ISSN: 2224-350X
Volume 14, 2019
A predictive convolutional neural network model for source-load forecasting in smart grids
Authors: ,
Abstract: Smart grid engineering is the key for an optimized use of extensive energy resources which allows hybrid renewable energy sources (RES) to be optimally integrated, and their power generation efficiently dispatched to the grid over long distance DC transmission lines using the high voltage DC (HVDC) transmission technologies. In this context, it is crucial to determine the required number of generating units of wind turbines and photovoltaic arrays, and the associated storage capacity for standalone and/or grid connected hybrid microgrid. Typically, this is determined using a sizing algorithm based on the observation that the state of charge of battery power system should be periodically invariant in order to keep optimum cost. Given the intermittency of wind speed and solar irradiance, it is very challenging to accurately calibrate power production from wind turbines and photovoltaic arrays, at all times even under varying weather conditions. In this context, the convolution neural network (CNN) can symbolize a practical and reliable tool to precisely monitor and predict the wind speed and solar irradiance outputs and accordingly manage the power transfer switching between areas that have surplus of renewable energies to areas with energy shortage by initiating the energy management system (EMS) to dispatch power according to an anticipated schedule. The efficiency of the proposed CNN based model was tested using meteorological data related to Beirut city. The experimental results indicate that the mean absolute error related to wind speed and solar irradiance are low, demonstrating very high forecasting accuracy.
Search Articles
Keywords: Smart grid, convolution neural network, energy management system, wind energy forecasting, solar energy forecasting.
Pages: 181-189
WSEAS Transactions on Power Systems, ISSN / E-ISSN: 1790-5060 / 2224-350X, Volume 14, 2019, Art. #22