Research On Water Consumption Data Collection And Water Event Identification In College Dormitories

With the continuous development of our education level,the number of college students is increasing,and colleges have become the major water users in China.Although the propaganda of water saving is quite effectually,students in colleges still have many problems in water saving.At the same time,with the popularity of intelligent devices,some college students’ daily work and rest become irregular,which seriously affects the growth of college students.Although most colleges have implemented the lights out system,it cannot fundamentally solve the problem of students staying up late.Therefore,this paper designed a neural network model that uses the water consumption data of college dormitories to identify the current water events.Based on the water consumption data and the identified water events and other related information,it can judge whether students have irregular work and rest and waste water,so as to strengthen the management of college students.At the same time,considering the accuracy and real-time performance of the collected water data,low power consumption and low cost of the collection device,this paper designs a water consumption data acquisition device with a low power microprocessor as the core to realize realtime collection of water consumption data in college dormitories and accurate acquisition of water consumption data.The main work of this paper is as follows:(1)In view of the shortcomings of manual reading of water consumption data,such as unguaranteed real-time performance and incomplete data collection,this paper uses low-power microprocessor STM32F103RCT6 as the core,a pulse sensing device composed of magnetic pointer and reed tube,and LORA as a communication device to design a water consumption data acquisition device.The installation of this device on the basis of the original water meter can realize the real-time collection of water consumption data in college dormitories and meet the requirements of low cost,low power consumption and small volume.(2)BP neural network and CART decision tree algorithm were used to identify water events collected from college dormitory water use data,and then compared with actual water events.The results show that the BP neural network has a higher recognition accuracy than the CART decision tree model,so the BP neural network is used as the base classifier in this paper.(3)Researches on the identification of water events are relatively limited,and most of them only involve a single water event,lacking researches on more complex simultaneous water use situations.In view of the recognition error caused by the same water use,this paper puts forward a RAKEL-BP multi label classification algorithm based on improved SMOTE to identify the water event,and compare with the traditional BP-MLL,LP-BP,RAKEL-BP algorithms.The results show that the RAKEL-BP algorithm based on improving SMOTE has a good performance in various evaluation indexes,and can effectively improve the accuracy of recognition of water event.(4)According to the start time of water use and the current water event identified,it can be judged whether students carry out water use behaviors such as washing and bathing at a later time,so as to determine whether there is the phenomenon of staying up late.At the same time,in view of the water saving problem,the phenomenon of water waste can be identified by setting excess water consumption to strengthen the water-saving work in colleges for the water events that are deeply affected by human factors such as washing and bathing.