The spatial load forecasting（SLF） of old and new urban areas without historical load data commonly uses an average classification load density based approach， which is called classification and subarea method to forecast the small areas load of similar kinds. Based on vacant area inference（VAI）， the classification and subarea method is improved. By subtracting the classification load of old city areas with historical load data from the classification load of existing areas（not including new city areas）， the result is the classification load of old city areas without historical load data. Similarly， by subtracting the classification load of existing areas from the classification load of planning city areas（including new city areas）， the result is the classification load of new city areas. And then， by calculating the average load density of new and old city areas without historical load data combining with their area， the loads of all areas are forecasted. The method proposed improves the forecasting accuracy because different thinking lines are employed for new city areas and old city areas with and without historical load data， but little additional data are needed. The applicability and effectiveness of the method are demonstrated by using a real SLF case.