The table illustrates that the model result fits the historical power consumption data with an average coefficient of a high multiple determination (R2-value). Testing the model by F-test, a validation for the significance of all regression coefficients, we can draw the conclusion that all factors have remarkable effects on the building power consumption. Using t-test, which represents the significance of a single variable, we can see that the variable DAY of commercial buildings is not significantly correlated with the building power consumption. An acceptable reason is that the influence of weekend is insignificant on the power consumption though customers and sales surge at weekends. So we redo the regression process after eliminating the factor-DAY. The result is presented in Table 3.
Table 3 Modified regression results of total power consumption in commercial buildings
Type CB1
CB2
Season Cooling weeks
Cooling weeks
Impact factor C
CDH C CDH
Regression coefficient 170720.9 47.53724 296627.7 69.8585
F-test 5.58E-22
1.4E-23
t-test R2 1.99E-31
5.58E-22 0.95 2.28E-35
1.40E-23 0.96
As seen from the table, the fitness of the model is still good. For there is only one factor left, F-test and t-test are the same, indicating that the CDH has a remarkable influence on the power consumption of the commercial building.
We use the trained prediction model to predict the power consumption in 2014 and validate the prediction effect by real-time monitoring data. Table 4 demonstrates the predictive effect of total power consumption of different buildings types in cooling season.