Hello to all,
I'm Arianit Zeqiri a student at the University of Prishtina.
I'm preparing a research work to predict the global solar radiation by classical methods and to compare it with methods of AI ( ANNs).
I have used data from about 340 cities (together), where the data for each city are in daily average monthly value where are measured data for about 10 years (1966-1975).
I have got very good results of R-value from a classical method (about 96%), now I decide to use ANN methods to compare it with the classical method. I have used the nftool in MATLAB and set as INPUT the (S/S0), Somax, Kmin, Kmax, Gmin, Gmax, Latitude, Longitude, and Evaluation and as TARGET the (H/H0), where I have become a result of R-value as about 98 %.
P.S in nftool I have divided the data in 70/15/15 and used 5 Hidden Layers, 1 output layer.
My questions are:
What tells us this output (predicted) data, are they the future predicted GSR or what?
What is the impact of predicting GSR with ANN?
Somewhere I have read that ANNs create a relationship between input data with the output, so when this stands, what is the equation that comes from ANNs?
I would please you to tell me how to use ANN for this case and how to present those predicted data.
Hope anyone will answer my questions, or to create a relationship to work together in this case.
Thanks to all.
Hi Arianit Zeqiri,
before you dive into using ANN for prediction, you need to first understand the state-of-the-art in solar irradiance forecasting. After reading what you have described, my impression is that your data is out-dated and its resolution is too low.
ANN has been well-explored for solar forecasting (thousands of papers available, maybe more), so I think just applying MATLAB functions is not going to lead you very far. I would suggest you reading the literature. For instance, I attached a very recent review article by five editors of Solar Energy Journal (which is one of the most popular journals for solar forecasting works). If you have any questions, do not hesitate to contact me.
cheers
dazhi
Hi Arianit Zeqiri,
before you dive into using ANN for prediction, you need to first understand the state-of-the-art in solar irradiance forecasting. After reading what you have described, my impression is that your data is out-dated and its resolution is too low.
ANN has been well-explored for solar forecasting (thousands of papers available, maybe more), so I think just applying MATLAB functions is not going to lead you very far. I would suggest you reading the literature. For instance, I attached a very recent review article by five editors of Solar Energy Journal (which is one of the most popular journals for solar forecasting works). If you have any questions, do not hesitate to contact me.
cheers
dazhi
have a look at this
Conference Paper MATLAB tool for predicting the global solar radiation in UAE
you can use nftool it is the best way to do it
https://www.mathworks.com/help/nnet/ref/nftool.html
There is not much value in predicting monthly-mean irradiance. Moreover, sunshine measurement is imprecise by nature. It has been abandoned decades ago in various countries for that reason. The current best practice is to use observations from geosynchronous satellites. I'm afraid that your project is not going in the right direction...
I agree with Chris Gueymard, the only predictable monthly mean solar irradiance in the cloudless sky conditions (clear sky), if you have good aerosol and ozone climatologic models. Extrapolation based on long series of ground or satellite observations is the best way to estimate the possible monthly-mean solar irradiance.
I hope these remarks could help you
Didier Gillotay
Hi eng. Zeqiri
I would like inform you that I am working on this project , please could you provide me the matlab code or related information?
best regards
- Badges
- Science topic