Novel Hybrid Machine Learning Algorithms for Lakes Evaporation and Power Production using Floating Semitransparent Polymer Solar Cells

Abstract

AbstractThe present study predicts the future evaporation losses by applying novel hybrid Machine Learning Algorithms (MLA). Water resources management is achieved by covering the reservoir water surface with floating semitransparent polymer solar cells. The energy produced by these panels will be used in the irrigation activities. The study is applied for the mass water body of Nasser Lake, Egypt and Sudan. Five MLAs namely additive regression (AR), AR-random subspace (AR-RSS), AR-M5Pruned (AR-M5P), AR-reduced error pruning tree (AR-REPTree), and AR- support vector machine (AR-SVM) were developed and evaluated for predicting future evaporation losses in the years 2030, 2050, and 2070. The study concludes that the hybrid AR-M5P ML model was not only superior to the AR model alone but also outperformed other hybrid models such as AR-RSS and AR-REPTree. The expected total annual water saving are projected to reach 3.47 billion cubic meters (BCM), 3.68 and 3.90 BCM, while the total annual power production is observed to be 1389 × 109 Megawatt (MW), 1535 × 109 MW and 1795 × 109 MW in the years 2030, 2050 and 2070, respectively. These results were achieved by covering the shallow water depths from contour level 0 m to 10 m below the surface water level. Additionally, this study shows the ability of using MLAs in the estimation of reservoir evaporation and addressing the water shortages in high stress regions. Graphical Abstract