ارزيابي شبيه سازي رواناب حوزه هاي برفي با مدل شبيه سازي (SRM) و شبكه عصبي براي برآورد انرژي برقابي در مواجهه با كمبود آمار

Runoff Simulation in Snowbound Catchments, Using SRM and ANN Models to Estimate Hydropower Potentials in Data Scarcity Situations

Small hydropower plants can have an important role in energy generation. Upper catchments are normally proper locations to construct such plants, but they usually have snowy regimes and the availability of data is usually a problem. This paper is an attempt to simulate snowmelt-runoff with SRM and ANNs in the relatively small catchment of Sardabrood in northern Iran with scarce data. In the next step, effects of errors resulting from the streamflow simulation on the estimated hydropower energy potentials is investigated. For the SRM model, a snow covered area is needed. This is met by the images of the A VHRR sensor of NOAA satellites for the years 1999 and 2000. In case of ANNs, the networks are trained with 1 year- (1999 using stations in the region) and 3 year- (1997 to 1999 using stations of the region and nearby) observed data. Year 2000 is used for verification. The results have shown that if ANNs get trained with 3-year data, it performs better than SRM. Both methods have problems in high flow simulations. Duration-Curve method and Sequential Streamflow Routing method are applied to simulate electricity generation, based on the results of runoff simulations. The RETScreen software and a program developed in this research are implemented for this purpose. The comparisons suggested better performance of SRM in the equal time periods (1999 and 2000) and subsequently better energy generation prediction. ANNs with 3 years training have closer results to SRM. Although runoff simulated with SRM have better performance in energy generation. This is because of better simulation of runoff in the operational ranges of the turbines.