Net load forecasting using different aggregation levels

Beichter, Maximilian; Phipps, Kaleb; Frysztacki, Martha Maria; Mikut, Ralf; Hagenmeyer, Veit; Ludwig, Nicole

doi:10.1186/s42162-022-00213-8

Energy Informatics

Table 1 An overview of the progressive feature selection for each target time series

From: Net load forecasting using different aggregation levels

	Target time series to be forecast
	Onshore wind	Offshore wind	Solar	Demand	Net load	Supply
Surface net solar radiation north Germany			14		34	32
Surface net solar radiation central Germany			9		12	14
Surface net solar radiation south Germany			2		2	2
Surface net thermal radiation north Germany			6		17	17
Surface net thermal radiation central Germany			3		27	20
Surface net thermal radiation south Germany			7		13	9
Temperature 2 m over ground north Germany			10		28	8
Temperature 2 m over ground central Germany			12		14	26
Temperature 2 m over ground south Germany			8		10	7
Eastward wind velocity Baltic sea		6			24	11
Eastward wind velocity North sea		4			31	31
Northward wind velocity Baltic sea		3			32	27
Northward wind velocity North sea		8			30	29
Resulting wind velocity Baltic sea		2			9	6
Resulting wind velocity North sea		2			6	2
Eastward wind velocity north Germany	5				35	24
Eastward wind velocity central Germany	13				18	18
Eastward wind velocity south Germany	4				11	13
Northward wind velocity north Germany	8				36	12
Northward wind velocity central Germany	9				20	23
Northward wind velocity south Germany	6				26	19
Resulting wind velocity north Germany	2				2	4
Resulting wind velocity central Germany	2				3	3
Resulting wind velocity south Germany	3				8	5
One hot encoded weekdays				5	5
One hot encoded hours	18	15	5	11	7	16
Public holiday				3	15
Partial public holiday of Germany				12	37
Summertime	19	16	4	13	33	15
Historical values (lags) for the last 0 h	7^a	5^b	2^c	2^d	4^e	10^f
Historical values (lags) for the last 24 h	15^a	12^b	13^c	10^d	21^e	33^f
Historical values (lags) for the last 48 h	17^a	9^b	19^c	8^d	29^e	25^f
Historical values (lags) for the last 72 h	10^a	11^b	15^c	6^d	23^e	34^f
Historical values (lags) for the last 96 h	14^a	14^b	17^c	4^d	19^e	22^f
Historical values (lags) for the last 120 h	12^a	7^b	11^c	7^d	25^e	30^f
Historical values (lags) for the last 144 h	16^a	10^b	16 ^c	2^d	16^e	28^f
Historical values (lags) for the last 168 h	11^a	13^b	18^c	9^d	22^e	21^f

Starting with a large number of features we iteratively reduce the number of features that can be selected. The number in each row indicates the smallest subset of features that still included the given feature. We perform feature selection until a minimal feature set size of 2
^aFor onshore wind, the lags correspond to historical onshore wind values
^bFor offshore wind, the lags correspond to historical offshore wind values
^cFor solar, the lags correspond to historical solar values
^dFor demand the lags correspond to historical demand values
^eFor net load, the lags correspond to historical net load values
^fFor supply, the lags correspond to historical supply values

Back to article page