A Novel Solar Linear Fresnel Reflector Field in Al-Khobar, Saudi Arabia: Experimental Investigation Assisted with Relevance Vector Machine Modeling and Simulated Annealing Optimization

This research introduces a detailed experimental investigation and advanced artificial intelligence modelling on a solar linear Fresnel reflector system (SLFRS), producing heat for industrial processes. The studied SLFRS features an inverted flat-plate receiver with a gross collection area of 306 m2, connected to a 5.0 m3 Therminol-66-oil tank. The system is analysed in terms of heat power capacity, exit oil temperature, and system energetic and exergetic efficiencies. The operational performance of the SLFRS is modelled by relevance-vector machine (RVM). An enhanced version of simulated annealing optimization algorithm is combined with RVM to boost prediction accuracy. Experimental findings showed that the designated SLFRS could generate 62.38 kW thermal power, with energic and exergic efficiencies of 35.11% and 7.84%, respectively. The system also proved capable of achieving temperatures up to 213.35 °C, at 9.0 m3/h thermal-oil flowrate. Statistical evaluations highlighted the RVM-SAOA model's effectiveness over competing models in performance prediction, showing R2 of 0.9997 and 0.9988, and minimal RMSE values of 2.5150 and 0.4091 for predicting the exit therminol-oil temperature and system's heat power, respectively.