Solar intermittency is a major issue for solar photovoltaic (PV) power plants as the unpredictability of energy supply can disrupt grid integration with renewables. Solar forecasting is a solution for grid operators to predict and balance energy supply and demand. PV performance is affected by the global horizontal index (GHI), which measures the radiation received per unit area on the earth's surface that has been scattered by atmospheric and weather conditions. A statistical model forecast using a dataset containing meteorological information such as GHI, ambient temperature, air pressure and relative humidity is used to predict PV performance. A pre-processing step to embed time awareness is necessary to produce a high-quality model. An artificial neural network (ANN) with radial basis functions (RBFs) as an activation function is used to forecast GHI by analyzing historical data on GHI and meteorological parameters to identify non-linear relationships. This experiment aims to create a model to predict GHI in the upcoming 24-h period using 5 years of GHI data from NREL's national solar radiation database (NSRDB) as both training and test data. The model will be evaluated using local existing solar power plant data in the same area.
The aim of this project is to improve the current capability of solar forecasting model developed by PT Faraday Tekno Energi (Faraday). Faraday will provide the solar irradiance forecasting model for free with its open access policy.
Call for collaboration
We are currently using NSRDB open access GHI data to train and test the model. We are looking for collaborators who can provide us with local solar power plant data to evaluate the model. The data should contain the following information:
-
In any format:
- Location (preferrably within the mainland Java or Republic of Indonesia) Preferrably in .csv, .xlsx, or .json format:
- UTC-formatted timecode/ locally-formatted standardized timecode (WIB/ WITA/ WIT, etc.), logged at least every 10 minutes
- Time series of PV power output
- Local temperature (optional, preferrable)
- Local humidity (optional, preferrable)
The data is usually available from your inverter or monitoring system. If you are interested in collaborating, please contact us at co-write@haposan.com or click the button on top right. We will send you details alongside a formal letter to request your data.
In return, you will get access to our fully trained solar forecasting model earlier than the open-access public release. If you prefer to, your name will be included in the upcoming scientific publication related to this work.
If you have more questions about collaboration, co-writing, or researching together, please contact us at co-write@haposan.com.