Mini-split Air Conditioning System, Multi-variable, Newton Based, Extremum Seeking Control, Real Time Optimization
In this study, a multi-variable extremum seeking control (ESC) scheme is proposed for a variable-speed mini-split air-conditioning system. The control inputs are the evaporator and condenser fan speeds, respectively, while the total power consumption is used as the feedback. As accurate model is hard and expensive to obtain for the AC system of interest in real time, nearly model-free self-optimizing control methods such as ESC is considered a more feasible solution to practical deployment. Recent development in ESC, and especially the Newton based multi-variable ESC method with online Hessian estimation provides the capability for real-time decoupling among the input channels (Ghaffari et al. 2012). Different from gradient based multi-variable ESC method, the Newton based multi-variable ESC provides uniform convergence characteristics for all the control inputs. Therefore, the Newton based multi-variable ESC is suitable for multi-input real-time optimization, especially for the case with large gain variation and coupling for different control input channels. An experimental setup is developed with a 9000 BTU variable-speed mini-split AC system (Mitsubishi MSZ-GE09NA & MUY-GE09NA). A 2000 Watts electrical heater works as the heat load. The indoor unit of the mini-split system and the heater are installed in a 4’x8’x6’ insulated chamber. A Watt Node Pulse WNB-3D-240-P power meter is utilized to measure the power consumption of the mini-split system. To achieve the speed control of the evaporator fan motor and condenser fan, a TMS320F28035 based customized motor controller is used. Three RTD temperature sensors are deployed to measure the indoor temperature, the outdoor temperature and the condenser coil temperature, respectively. The data acquisition and control algorithms are implemented on a National Instruments CompactRIO platform. During the system operation, the CompactRIO reads the power consumption sent from the power meter, which will be fed into the ESC control algorithm to get the speed reference for both the evaporator fan and the condenser fan. Then, the speed reference will be applied to the motor controllers for each motor. Meanwhile, some other measurements such as indoor temperature, outdoor temperature, the speed feedback for both the motors, etc. are also monitored by the CompactRIO. The experimental study is planned to include three scenarios of ESC implementation: 1) single-input ESC with evaporator fan speed input only; 2) single-input ESC with condenser fan speed input only; and 3) multi-variable ESC with both evaporator and condenser fan speed inputs. Experimental study has performed for the first scenario. Under the ambient temperature of 75F and indoor room temperature set-point of 68F, the ESC control results in an energy saving of 20%. The work under way includes the other two scenarios and in particular the multi-variable ESC. More experiments will be performed under various weather conditions.