Design of an Artificial Neural Network Controller for a Tankless Water Heater By Using a Low-Profile Embedded System

Tankless water heaters (TWHs) have been become more popular day-by-day in special because of the low-power consumption that characterizes these devices in comparison with the tank water heaters. Nonetheless, it is desirable that these systems have a rapid response to disturbances such as changes in water flow or the inlet temperature. Different methods of classic control have been used for solving this problem for decades. These techniques provide a good solution although not necessarily the optimal one. With the recent boom in automatic control techniques based on Artificial Neural Networks (ANNs) [1]–[3] and the scaling in terms of computational power of embedded systems, this has led to the use of ANNs in low-profile embedded systems. In this work, we present an implementation of an ANN for a commercial application of a TWH running on a low-profile embedded system where we demonstrated that the stabilization time is reduced by up to 25% whilst the overshoot by up to 50%, both in comparison with a classic methods of automatic control using a low-performance microcontroller.