Performance optimization of the parabolic trough power plant using a dual-stage ensemble algorithm

The flexibility of future power production systems must be maximized in order to offset the unpredictability of non-dispatchable energy from renewable sources. The absence of state policies and strategies to encourage investment in exploiting solar energy is why it is not widely used in Africa. In light of this, this research first aims to carry out a techno-economic assessment of parabolic trough power (PTP) plants in six cities in Egypt. For this purpose, a 100 MW nameplate capability has been simulated using the system advisor model simulation environment. Finally, four machine learning models are proposed, including artificial neural network, Gaussian process regression, regression neural network, and least square boosting in conjunction with a generalized additive model (GAM) as a meta-model, in order to develop a generalized model to predict the PTP performance based on the data of ten different cities at Africa. Utilizing these longstanding machine learning models for feature extraction, EnsGAM is tailored to the optimal predictors The findings indicate that, with a capacity factor of 55.4 % and an annual energy output of 484.7 GWh, the Benban location produces the most energy. In addition, Benban exhibits the shortest simple payback period—10.1 years—while Kuraymat displays the longest—11.5 years. The findings showed that EnsGAM performs noticeably better than all comparison techniques, producing the highest correlation coefficients/Willmott’s agreement index for power generation and maximum discharge energy of 0.9463/0.9724 and 0.958/0.9778, respectively.