Solar Energy Forecasting in Vlora using Artificial Neural Networks and Open Data

Session

Computer Science and Communication Engineering

Description

Solar energy forecasting is considered an essential scientific aspect in supporting efforts to integrate solar energy into electricity grids. This is because grid operators need to know how much solar energy the system is producing so they can optimally engage solar and other energy sources to balance demand and production.

Improving solar power forecasts will allow the electric grid to be more flexible and adapt to changing conditions. This will in return help minimize outages and the overall cost of service.

Artificial Neural Networks (ANN) are powerful tools for modeling and estimating Solar Energy even though they use few inputs. To train the networks, a dataset of daily meteorological time series for a period of 12.5 years (2010–2022) was collected for the city of Vlora by Weather Data Service Visual Crossing, and publicly accessible, was used. The meteorological parameters used to estimate the solar energy were the daily values of the maximum, minimum and average temperatures; relative humidity; daylight hours; precipitation; wind speed; solar radiation, weather description as inputs. The output is a daily solar energy in MJ/m 2 day. Various ANN models have been designed and implemented by combining various meteorological data. The optimal model for estimating solar energy was an MLP with one hidden layer where the inputs were numerical values for maximum and minimum temperature, daylight hours, solar radiation, humidity, and weather description. The data used is Open Data, which makes the model very suitable to use for other regions as well.

To evaluate the difference between measured and predicted values by ANN models, mean absolute error (MAE), mean square error (MSE) and correlation coefficient (R) were determined. For the 6-5-1 topology - which is one of the best topologies - the R, MSE, and MAE values were found to be 0.999, 0.019, and 0.0868, respectively.

The obtained results showed that the ANN model can be successfully used to estimate the daily solar energy for Vlora and other locations.

Keywords:

Solar radiation; Solar Energy; Artificial Neural Networks; Forecasting; Open Data;

Proceedings Editor

Edmond Hajrizi

ISBN

978-9951-550-50-5

Location

UBT Kampus, Lipjan

Start Date

29-10-2022 12:00 AM

End Date

30-10-2022 12:00 AM

DOI

10.33107/ubt-ic.2022.265

This document is currently not available here.

Share

COinS
 
Oct 29th, 12:00 AM Oct 30th, 12:00 AM

Solar Energy Forecasting in Vlora using Artificial Neural Networks and Open Data

UBT Kampus, Lipjan

Solar energy forecasting is considered an essential scientific aspect in supporting efforts to integrate solar energy into electricity grids. This is because grid operators need to know how much solar energy the system is producing so they can optimally engage solar and other energy sources to balance demand and production.

Improving solar power forecasts will allow the electric grid to be more flexible and adapt to changing conditions. This will in return help minimize outages and the overall cost of service.

Artificial Neural Networks (ANN) are powerful tools for modeling and estimating Solar Energy even though they use few inputs. To train the networks, a dataset of daily meteorological time series for a period of 12.5 years (2010–2022) was collected for the city of Vlora by Weather Data Service Visual Crossing, and publicly accessible, was used. The meteorological parameters used to estimate the solar energy were the daily values of the maximum, minimum and average temperatures; relative humidity; daylight hours; precipitation; wind speed; solar radiation, weather description as inputs. The output is a daily solar energy in MJ/m 2 day. Various ANN models have been designed and implemented by combining various meteorological data. The optimal model for estimating solar energy was an MLP with one hidden layer where the inputs were numerical values for maximum and minimum temperature, daylight hours, solar radiation, humidity, and weather description. The data used is Open Data, which makes the model very suitable to use for other regions as well.

To evaluate the difference between measured and predicted values by ANN models, mean absolute error (MAE), mean square error (MSE) and correlation coefficient (R) were determined. For the 6-5-1 topology - which is one of the best topologies - the R, MSE, and MAE values were found to be 0.999, 0.019, and 0.0868, respectively.

The obtained results showed that the ANN model can be successfully used to estimate the daily solar energy for Vlora and other locations.