Data Cleaning

The data cleaning process took six months, commencing in early March 2009 and was completed in late September 2009. As up to seventy sensors were installed in each house, downloading data at 10 minute intervals, the number of files to be cleaned was significant and time-consuming. The data cleaning methods followed the same process as adopted for the test cells in Launceston and were developed in close consultation with CSIRO researchers (Dewsbury 2011). Starting with the raw base data (as acquisitioned from the data logger), a new version of the database was established for each checking and cleaning procedure. For the checking and cleaning step one (V1), the data was checked to see whether or not the temperature was within a predetermined range of measurements. The range of measurements was selected based on an educated assumption of expected measurements within the houses. Data outside the pre-determined range was subjected to further examination and was either adjusted or deleted as appropriate. During the last checking step, version 7 (V7), the data was checked for any unexplainable, unusual, or drastic shifts, fluctuations or patterns.

The actual data checking was undertaken by employees at the School of Architecture and Design’s Centre for Sustainable Architecture with Wood (CSAW). The project researcher was not involved in the data checking process, to allow for independent and objective checking, without the influence of the researcher’s personal engagement and intimate knowledge of the thermal performance of the houses. All enquiries by the data checking team were attended to and their outputs were checked by the project researcher. Technical experts from the University’s School of Engineering and Architecture and Design were engaged to assist with the setting of realistic range and step measurements for all sensor locations in the houses. Additionally, data checking involved cross-comparison, either with data from a nearby sensor or from additional relevant sensors in the weather station. Table 6.3 shows the step-by-step data cleaning process used for this project.

Table 6.4: Method of Data Cleaning of the Test Houses

V1 10 Minute Data Range Checks Each sensor device was allocated an expected range of

measurement. All the data were checked to ensure the measurements were within that range.

V2 10 Minute Data Step Checks Each sensor device was allocated an expected step value

within a 10 minute data reading. All data were then checked to ensure step measurements were within the pre-determined step check range.

V3 10 Minute Data Graphical

Checks

Graphical software converted the data into a graphical format. This analysis checked for abnormal shifts or unusual data patterns. Large data swings were analysed and checked.

V4 Averaging 10 Minute Data

into Average Hourly Format

The six individual 10 minute reading were averaged to an hourly value. The only exception was the averaging of wind speed and wind direction, which used a different method of establishing hourly values.

V5 Average Hourly Data Range

Checks

Each sensor device was allocated an expected range of measurement. All data were checked to ensure the measurements were within that range.

V6 Average Hourly Step Checks Each sensor device was allocated with an expected step

value within an hourly data reading. All data were then checked

V7 Average Hourly Data Graphical Checks

The final checking process was the application of graphical software to convert the data into graphical presentation. This method highlighted abnormal shifts or unusual data patterns.

V8 Average Hourly Data Specific selected data averaged and used for the empirical

validation with AccuRate

In data range step version 1 (V1) a ten-minute range check measurement was allocated to each sensor. This included some investigation and predetermining of realistic environmental measurement fluctuations for both inside and outside sensors. For example, the estimated expected inside roof temperature fluctuation in the 4-star timber floor house was between -3ºC and 50ºC, while the actual measured temperature range was only 0ºC to 35ºC. The project researcher had to accept or reject the measured temperature range, in consultation with experts in the Schools of Architecture and Design and Engineering. In general terms, the estimated environmental measurements of range and step values were within the range of actual measured values.

The graphical checking of data required a different mechanism. Here, a ten minute and an hourly interval data were checked and analysed. Unusual fluctuations, changes in patterns, or drastic spikes and sharp dips, were investigated to determine the validity of the measurements. This often required the checking of other nearby sensors, or the additional investigation of weather patterns occurring at that time. For example, some sensors showed drastic spikes in temperature and only after some further research was it realised that the sensor had been exposed to solar radiation at that particular time. These phenomena occurred in the early mornings to the sensors located in the bedrooms and in the late afternoon to the sensors located in the living room. Figure 6.66 shows a sample of a graphical presentation of temperature gradient in the living room of the 5-star timber floor house during a 5-day period, based on averaged hourly temperature values.

Figure 6.66: Temperature graph for a 5 day interval in the centre of the living room at 1200 height from floor level. Here the dip in temperature on day 2 was investigated

As mentioned above, a sharp dip in temperature at 9 a.m. was questioned by the data checking team. Further examination of the trend of solar radiation showed a distinct dip in the amount of solar radiation at the eastern wall at exactly the same time as the temperature dip in the living room occurred. Figure 6.67 shows the fluctuation of solar radiation at the eastern wall, which explains the pronounced swings in temperature measurements.

Solar Radiation Eastern Wall Vertical kWh/m² 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 8:00 9:00 10:00 11:00 12:00 13:00 18.7.2007 18.7.2007 18.7.2007 18.7.2007 18.7.2007 18.7.2007 Time S ol ar R adi at ion (kW h /m 2

Solar Radiation Eastern Wall Vertical kWh/m²

Figure 6.67: Solar radiation received at the eastern wall of the 5-star timber floor house on 18.7.2007 between 8 a.m. and 1 p.m.

The graphical checking of the data and analysis of the unusual temperature fluctuations, spikes and dips showed that all the data was useable, once the reasons for the unusual temperature profiles were fully understood.

6.7. Summary

An extensive range of measurements was collected during the three months of free-running operation of the houses. While only a limited number of measurements was necessary for this validation process, many additional measurements were taken, to provide valuable data backup as well as data for future thermal performance research outside the scope of this study.

Utmost care was taken to ensure the proper installation of cables, plug connectors, sensors and all other measuring devices. Testing of wiring, cables and sensors included an elaborate checking of calibration and mechanisms to ensure the proper functioning of the system. For the purpose of this empirical validation study, environmental measurements for three weeks duration (from 5 September to 26 September) were processed rigorously into a complete data set for comparison with the thermal performance of the houses as simulated by AccuRate. The relevant data files were:

 Roof space dry bulb air temperature;

 1800mm height at centre of room dry bulb temperature for all rooms;

 600mm height at centre of room dry bulb air temperature for all rooms;

 Subfloor dry bulb air temperature in the timber floor houses;

 All weather station measurements.

The reliable instrumentation and meticulous data checking and cleaning mechanism resulted in a high quality data set for the empirical validation of the HERS AccuRate. The Chapter 7 focuses on the simulation software AccuRate and describes the appropriate simulation procedures for the empirical validation of AccuRate.

Chapter 7 also illustrates the preparation of AccuRate’s input data required for the empirical validation process and explains the thermal simulations of AccuRate.

Chapter 7: AccuRate Thermal Performance Simulations of