Fuzzy logic and neural networks based solar radiation prediction

Predictions of incoming solar energy are acquiring more importance, because of strong increment of solar power generation. Predictions is very useful in solar energy applications because it permits to generate solar data for locations where measurements are not available. In existing systems, solar radiation is predicted using fuzzy logic and neural networks separately. So that Mean absolute percentage error is greater than 10%. In our proposed method, Fuzzy logic and neural networks are combined together using Takagi Sugeno Kang (TSK) method. TSK method is very efficient than mamdani method. Previous year solar radiation data is collected from National Environmental Agency and using this values neural networks was trained. The graph between measured and predicted data values was plotted. Error is calculated using the difference between desired and output value. Prediction using combination of fuzzy and neural network model having Mean Absolute Percentage Error (MAPE) is less than 10%.So that this method will reduce the mean absolute percentage error is much smaller compared with that of the other solar radiation method