Temperature and Humidity Dataset of an East-Facing South African Greenhouse Tunnel
Description
The aim of this study was to model the thermal characteristics of an east-facing greenhouse tunnel in South Africa. The tunnel used has a fan and wet wall on opposite ends and is used to cool the tunnel when temperatures reach a certain threshold. In this study, the authors' research hypothesis was the development of accurate empirical and analytical thermal models that would be validated using the available measured data. Three temperature and humidity sensors were placed near the fan, the wet wall, and in the middle of the 29-meter-long tunnel. The middle sensor's temperature measurements were used to control when the fan turned on and off. When the temperature reached above 30 degrees Celcius the fan was turned on, and when the temperature reached 22 degrees Celcius the fan was turned off. This created an 8-degree Celcius hysteresis band in which the fan and wet wall was controlled. The ``Training Set.csv`` file was used to train a Support Vector Regression (SVR) model to predict and simulate temperatures an hour in advance using previous predictions. ''Test Set.csv'' was used to validate and measure the accuracy of the SVR model. Also, the analytical model's accuracy was measured using the ``Test Set.csv`` file to quantitatively compare the two models. The test set and training set encompasses 42 days of measurements, aggregated with solar radiation and ambient temperature measurements that were captured by MeteoBlue (www.meteoblue.com). As this weather data was in an hourly format and the three temperature and humidity sensor data was in 5-minute intervals, the data from MeteoBlue was linearly interpolated for improved model training. For the ``Full Data Set.csv``, the data includes 162 days of only the three sensor's temperature and humidity measurements, and the fan and wet wall state.
Files
Steps to reproduce
For the three temperature and humidity sensors, the DHT22 sensor was used with a Raspberry Pi. The Raspberry Pi measured each sensor every minute and made an average after 5 minutes. The middle sensor was measured every 10 seconds with the median value being used after 1 minute to control the fan and was averaged after 5 minutes. The data was uploaded to Dropbox (www.dropbox.com) and downloaded at a later stage for analysis. The fan and wet wall were controlled with a Solid State Relay (SSR), which controlled the neutral wires of the fan and wet wall that operated through 3-phase magnetic contactor. Python was used to analyse the data and create both models.