TY - GEN
T1 - Solar power prediction using deduced feature of visibility index and artificial neural network
AU - Shubham,
AU - Padmanabh, Kumar
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/11/30
Y1 - 2017/11/30
N2 - Renewable source of energy has lots of relevance today. Solar energy is available in abundance and it is preferable because it is pollution free and there is no cost attached to it. However solar energy is not reliable in nature because its production depends upon weather condition, time of the day and other local parameters. If a consumer have accurate prediction of solar power, the trade-off with other sources of energy could be done easily to minimise total cost of energy. Therefore accuracy of prediction of solar power is a key factor in planning and management of overall energy requirement. This paper presents mechanism of forecasting of solar power using artificial neural network (ANN).The used Multilayer Perceptron (MLP) model takes calculated extraterrestrial solar irradiation, meteorological forecasts of cloud conditions, humidity, visibility and temperature at the deployment site as inputs and has single neuron output. The main contribution of this paper is the definition of unique feature vector such as visibility index derived from public weather data. It is argued that such weather forecast is publically available which could be used to deduce solar power that could be transferred to the grid. The model has been developed on the private data obtained from our own deployment in Lechfeld, Germany and on public data available at Mercury, Nevada, USA. The first model predicts direct solar output delivered to the grid whereas the second model predicts Global Horizontal Irradiation (GHI). The first model gives average accuracy of 85%-89% whereas the second model gives an accuracy of 92%.
AB - Renewable source of energy has lots of relevance today. Solar energy is available in abundance and it is preferable because it is pollution free and there is no cost attached to it. However solar energy is not reliable in nature because its production depends upon weather condition, time of the day and other local parameters. If a consumer have accurate prediction of solar power, the trade-off with other sources of energy could be done easily to minimise total cost of energy. Therefore accuracy of prediction of solar power is a key factor in planning and management of overall energy requirement. This paper presents mechanism of forecasting of solar power using artificial neural network (ANN).The used Multilayer Perceptron (MLP) model takes calculated extraterrestrial solar irradiation, meteorological forecasts of cloud conditions, humidity, visibility and temperature at the deployment site as inputs and has single neuron output. The main contribution of this paper is the definition of unique feature vector such as visibility index derived from public weather data. It is argued that such weather forecast is publically available which could be used to deduce solar power that could be transferred to the grid. The model has been developed on the private data obtained from our own deployment in Lechfeld, Germany and on public data available at Mercury, Nevada, USA. The first model predicts direct solar output delivered to the grid whereas the second model predicts Global Horizontal Irradiation (GHI). The first model gives average accuracy of 85%-89% whereas the second model gives an accuracy of 92%.
UR - http://www.scopus.com/inward/record.url?scp=85042745396&partnerID=8YFLogxK
U2 - 10.1109/ICACCI.2017.8125823
DO - 10.1109/ICACCI.2017.8125823
M3 - Conference contribution
AN - SCOPUS:85042745396
T3 - 2017 International Conference on Advances in Computing, Communications and Informatics, ICACCI 2017
SP - 97
EP - 102
BT - 2017 International Conference on Advances in Computing, Communications and Informatics, ICACCI 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2017 International Conference on Advances in Computing, Communications and Informatics, ICACCI 2017
Y2 - 13 September 2017 through 16 September 2017
ER -