Expert systems and the prospects of artificial intelligence for the automatic supervisory control of salinity gradient solar ponds

Salt Gradient Solar Ponds (SGSPs) are conventionally operated wholly manually. They hence suffer from a loss of efficiency for substantial durations when disturbed by natural diffusion and environmental elements such as temperature variation, solar radiation, and wind. They also impose considerable operational costs and time when a repair is unavoidable. Here, a Zero-Discharge Closed Cycle (ZDClC) is proposed for sustainable implementation of control architectures. We then define the SGSP as a control system and propose an automatic approach that uses Artificial Intelligence (AI) to mimic experts’ actions during repair operations. More specifically, an Automatic Supervisory Controller (ASC-SGSP) is proposed that decomposes the pond into several uniform sublayers according to their thicknesses and salt concentrations. The automated supervisor then uses a decision tree analysis of pond stability to determine the sequences of injection/suction actions to guarantee the pond stability and prevent layer overturning. An on–off subcontroller then directs the chosen sublayer's salt density to its desired value. The supervisory structure is general and can operate with either the proposed open-loop or closed-loop configurations. Results indicate the potential utility of AI and expert systems as a successful paradigm for SGSP control. Specifically, the proposed approach is examined in a simulated pond under several scenarios challenging its efficiency, adaptivity, robustness, and constraint handling properties. The ASC-SGSP fulfills the automatic control requirements during either open/closed-loop configurations and regulates the desired salinity levels under parameter constraints and uncertainties, even when the SGSP parameters are perturbed by as much as 50%.