Solar Power Forecasting and Planning Using Soft Computing Approach

Abstract

This thesis deals with the forecasting of solar power, it s planning and control using integrated soft computing approaches. The work is divided and conducted into three main parts. In the first part solar data is accumulated using DAQ from the panels deployed at the rooftop of Faculty of Engineering, Dayalbagh Educational Institute. Different variables which affect solar power generation are monitored and the data is collected. These variables include solar power, temperature, current, wind velocity, atmospheric pressure, and clouds. Solar power forecasting is performed by modelling two algorithms named Flexible Neural Network (FNN) and Genetic Algorithm Fuzzy Flexible Neural Network (GA-F-FNN) for different time horizons i.e., for 30 sec, 1 min, 10 min, and 15 min. using the above mentioned data. It is found that amongst these two algorithms GA-F-FNN showed better results as compared to FNN. Different factors affecting these two algorithms are also studied and monitored and finally forecasting of solar power is recorded. Also the effect of different factors like time horizons, static and rotating panel types, movement and percentage of clouds hindering solar panels, noise, normalization ranges etc. are studied during solar power forecasting. newlineIn the second stage planning of household appliances is done on the basis of solar power forecasting results obtained in the first part. Appliances are segregated based on power levels and priority levels and are bunched together accordingly. The classification of loads and their priorities are set for performing control in the final step. In the third and the final step, controlling of the planned appliances is done using fuzzy controller. The appliances are controlled on the basis of priority and the power level factors set for them. newlineAs a conclusion it can be said that accurate forecasting results help in better planning and controlling of appliances at domestic level. newline newline