An Intelligent Power Socket with Usage Behavior Prediction Through Frequent Pattern Trees

• Main Authors: LIAO, CI-REN, 廖啓荏
• Other Authors: CHI, KUANG-HUI
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/g8s9qh

碩士 === 國立雲林科技大學 === 電機工程系 === 106 === Although the advancement of technology and the introduction of electronic devices have made our life more and more convenient, demands for electricity become ever increasing in the meantime. In view of trends for saving energy, most current solutions deal with the problem by reducing the standby power consumption of electrical devices. Nowadays, smart sockets are commercially available, but they mostly function as timing switches or monitor dynamics of power consumption, without sufficient intelligence to operate in line with user’s habits of using electrical apparatus. As a result, such sockets require repeated manual setting or working with a master node, at expense of inconvenience or costly equipment. This thesis presents a hardware-software codesign based on frequent pattern trees which are leveraged to keep track of the usage behavior of home appliances. Our development embeds computing machinery that reduces user’s manual intervention wherever possible to maximize the intelligence of smart sockets. Our development consists of two parts, hardware and software. The former is implemented over Arduino with a current sensor, a solid-state relay and a Bluetooth module. Arduino is tasked to determine, from the readings reported from the current sensor, whether the plugged-in electrical device is presently operational, instruct the relay to switch on/off, and receive commands from the computer side. The software enables the user to select the matched electrical appliances, store data, monitor the status of the device, and predict whether or not the concerned device is correctly in use. The algorithm of frequent pattern trees is employed to analyze the stored records about when and what appliances were turned on/off in the past week, so as to find the usage pattern of certain devices. The constructed tree serves the purpose of forecasting user behavior. Our development operates automatically without manual intervention from the user. Currently our design is applicable to interactive devices such as LCD computer monitors, fans, lights that typically require switching on/off. Our mechanism resolves during which period of time each appliance probably remains off and its power provision can be cut off by the relay at the socket. Experimental results show that our prediction has higher precision when the user tends to use certain appliances regularly. Additionally, more power can be saved when the appliance remains off most of the time. Our design works fine along with user’s habits such that eventually a best suitable form of frequent patterns can be found for maximized utility, where longest hours of power saving for household appliances can be achieved with highest precision of prediction.