Establishment of Combining Grey Model with Partial Least Squares Regression for City Energy Consumption Forecasting

In this paper, a forecasting method, combining grey model (GM) with partial least squares regression (PLS), was applied to forecasting city terminal energy consumption for the first time. PLS is of the advantage of overcoming the correlative affect among independent variables, distinguishing the system signal from noises, and explaining dependent variables well; GM (1, 1) model is able to overcome the non-linear interference of parameters. The more accurate prediction could be achieved by the combination of these two methods to forecast the complicated and non-linear energy consumption system. The single model and combined model based on Dalian´s historical data were established, verified and compared. The result showed that relative err from the combined method was 2.93%, less than that from any method alone. From the prediction results by the combined method including seven variables, the city terminal energy consumption of Dalian were 18.76 Mtce, 29.85 Mtce, and 46.90 Mtce in 2010, 2015 and 2020, respectively, providing reference frame for energy planning and decision-making.