Morocco’s AI-Powered Hybrid System Redefines Hot Water Efficiency

In a quiet lab in El Jadida, Morocco, a team led by Mamhoud Redouane at the LabSIPE, National School of Applied Sciences, University Chouaib Doukkali, is quietly rewiring how buildings can access hot water. Their work, published in *E3S Web of Conferences* (translated: *Environmental, Energy and Earth Sciences Web of Conferences*), proposes a system that doesn’t just tap into the Earth’s warmth or the Sun’s rays—it orchestrates them, using a fusion of two renewable sources and a dash of artificial intelligence to predict exactly how much hot water is available at any given moment.

The idea is deceptively simple: combine surface geothermal energy—harnessing the relatively stable warmth just below ground—with solar thermal collectors, both circulating heat through a closed loop to a heat exchanger inside a domestic hot water (DHW) tank. The twist? Instead of relying on rigid thermostats and preset temperatures, the system uses a Physics-Informed Neural Network (PINN) to forecast hot water availability in real time, blending actual sensor data with thermal physics.

“Traditional systems maintain temperature at a set point, but they don’t adapt to actual demand,” explains Redouane. “With PINN, we’re not just reacting—we’re predicting. We can estimate how much hot water is stored and when it will be available, based on weather, usage patterns, and thermal dynamics. It’s like having a thermal crystal ball.”

For energy providers and building managers, this could mean fewer wasted kilowatt-hours, lower operational costs, and a stronger case for integrating hybrid renewable systems into urban infrastructure. The commercial implications are significant: buildings could reduce reliance on fossil-fueled boilers, cut peak demand charges, and even participate in energy flexibility markets by selling surplus renewable heat back to the grid.

While the technology is still in the simulation and early validation phase, the integration of PINNs represents a leap toward smarter, self-regulating thermal systems. Unlike black-box AI models, PINNs are constrained by physical laws—ensuring predictions are not just data-driven, but thermodynamically plausible.

As Redouane puts it, “We’re not replacing physics with algorithms. We’re using algorithms to better obey physics.”

The next step? Prototyping at scale. If successful, this architecture could become a blueprint for next-generation DHW systems in residential, commercial, and even district heating networks—especially in sun-rich, temperate climates where both solar and shallow geothermal resources are abundant. It’s not about reinventing the wheel, but about making sure the water in the tank is always just the right temperature—before anyone even turns on the tap.

Scroll to Top
×