and the human use [9]. The first few are considered as factors related to temperature and the last is with regards to the working duration.
Degree days (DD) represent a versatile climatic indicator frequently used in the analysis of building energy performance, which require less data input and can be used to assess rapidly how energy consumption may be influenced by major design decisions (e.g. insulation level, glazing area rate of the building, assumptions about infiltration, etc.) [10]. Heating degree days (HDDs) are defined as the deviation of the outdoor mean temperature from a heating reference temperature, taking into account only positive values. The reference temperature is also known as the base temperature which, for buildings, is a balance point temperature, that is, the outdoor temperature at which the heating systems do not need to run so as to maintain comfort conditions. Likewise, cooling degree days (CDDs) are calculated from temperatures above the base temperature. In this case, a base temperature is regarded as the outdoor temperature below which a building needs no cooling. However, in this paper, we introduce the cooling degree hours (CDHs) and heating degree hours (HDHs), similar concept of CDDs and HDDs owing to the fact that the precision of degree days is not sufficient enough for the short-term prediction. For a short period of time (daily, weekly, etc.), accumulated cooling/heating degree-hours (ACDHs/AHDHs) are calculated using the following expression:
ACDH N (CDH) ifTj Tb thenCDH Tm Tj (4) j 1 j elseCDH 0
AHDH N (HDHj) ifTj Tb thenHDH Tb Tj (5) j 1 else HDH 0
Where N is defined as the period of time i.e. number of hours in the week. The corresponding number of hours for the accumulated degree-hours for any period of time is determined by summation of the hours with the difference between Tj and Tb.
For office buildings and building complex, there exists an apparent difference in power consumption level between workdays and weekends. CDH, HDH and DAY are chosen as the impact factors in the regression model for weekly prediction, where DAY represents the number of workdays in a week. As for large commercial buildings, the flow of customers increases rapidly in the weekend, inevitably exerting great influence on the power consumption. In view of the huge inner zone and high internal heat, we choose CDH and DAY as the impact factors in the regression model, leaving the significance level of respective factors to be tested.
2.4. Season division
The whole year can be divided into cooling season, heating season and transition season. The base temperature discussed above cannot be determined arbitrarily or customarily. We can infer the base temperature from the historical power consumption data on the sub-metering platform. From Figure 1, it can be deducted that the commercial buildings begin cooling when the outside air temperature is below 15°C. For the office buildings and building complexes, air-conditioning system starts when the outdoor air temperature is over 20°C, and heating starts as long as the outdoor temperature is below 12°C. So for the office buildings and building complexes, the day is defined as the cooling day when Tm (daily mean temperature) is above 20°C. Likewise, the day is defined as the heating day when Tm is lower than 12°C, otherwise it is deemed as a transition day. For the commercial buildings, the day is defined as the cooling day when Tm is above 15°C, otherwise it is defined as a transition day. The detailed season division approach of a day based on Tm is tabulated in Table 1.