Household water consumption dataset

Published: 18-01-2019| Version 1 | DOI: 10.17632/2yjwrft6nr.1
Binoy Nair,
Sivasurya Kandasamy,
Rahul Balakrishna,
Bharathvajan L,
vamsi krishna Upputuri


The experimental setup was installed in a house with a Solar water heater (capacity 300 Liters) that gets filled every midnight at 12.00 onwards from Tank 1. Hence, the level changes logged for Tank 1 during the hours 00.00 to 03.00 am are high. During the rest of the day, the consumption is that of a regular household. The household has four adult occupants, with three occupants leaving the house by 10.00 am every weekday and returning by 4.00 pm. Tank 1 is used for supplying all household needs. Tank 2 is used only for gardening purposes. The house chosen for the experiment has two tanks. Sump is underground, from which the water is pumped to the overhead tanks Tank 1 and Tank 2. An Ultrasonic distance sensor is mounted near the top of each tank that monitors and logs the distance of the water surface from the sensor at fixed intervals. There are 13 sets of data for Tank 1 and 16 sets of data for Tank 2. The description reg. these datasets are given in Tank 1-Data Description.xlsx and Tank 2-Data Description.xlsx, respectively. The data logged is the vertical distance of the water surface in the tank from the sensor's location. The minimum reading of sensor is 19 cm which indicates full tank and the maximum reading is 88cm which indicates an empty tank. So to calculate the percentage of tank that has been filled as of now, we use the formula: percent_full=100-(100*(data-19)/(88-19))) The data is logged once very 15 seconds. The data logged is the one adjusted with calibration equation. Hence, the logged value is: level(t)=p1*sensor_reading(t)+p2 The equation found using OLS regression. Sensor calibration results: f(x) = p1*x + p2 Coefficients (with 95% confidence bounds): p1 = 1.033 (1.024, 1.042) p2 = 1.187 (0.7033, 1.67) Goodness of fit: SSE: 7.109 R-square: 0.9995 Adjusted R-square: 0.9995 RMSE: 0.4951