LAUSR

Abstract In complex outdoor conditions, radical weather changes can sometimes undermine the precision of temperature control systems, mainly because conventional heater controllers lack the ability to adapt to unpredictable parametric variations. In this paper, a heater auto-tuned by a PID neural network was proposed. Without knowing the range of weather variation in advance, the PID neural network self-adapts to weather changes and other kinds of disturbances, using a function that is driven by the back propagation algorithm. The temperature-control performance of this heater was numerically studied under a variety of outdoor conditions. A classical PID controlled heater was tuned under conditions as same as the PIDNN controller was pre-trained, and their performances were compared. The results showed that the PID neural network-controlled heater adapted well to weather and climate changes. It consistently maintained the temperature of the controlled unit with an overshoot of less than 0.2 °C, and it had a settling time of less than 32 s. By contrast, the PID controlled heater failed to achieve precise temperature-control when the wind speed rose at a rate greater than 1.5 m/s per hour. When the electrical resistance of the heater was temperature-dependent, the PIDNN controller managed to stabilize the temperature in less than 40 s. As for fast disturbances, such as sudden rain, the overshoot of the PIDNN was less than 1 °C, and the settling time was about 20 s.