Objective & Background
This paper provides a benchmark to evaluate operational day-ahead solar irradiance forecasts of Global Forecast System (GFS) for solar energy applications in China. First, GFS day-ahead solar irradiance forecasts are validated quantitatively with hourly observations at 17 first-class national solar monitoring stations operated by China Meteorological Administration (CMA) and 1 Baseline Surface Radiation Network (BSRN) station. Second, a hybrid forecast method based on Artificial Neural Network (ANN) and GFS product is proposed to improve forecasts accuracy, and it is validated against ground truth at Xianghe and Geermu. The results demonstrate parameter optimization of direct-diffuse separation fails to reduce the errors of direct normal irradiance (DNI) forecasts, and the hybrid method has the best performance for DNI forecasts.
Principal results
Tab. 1: Performance statistics of GFS day-ahead solar irradiance forecasts at 18 stations (unit: percent)
|
ID |
GHI rMAE |
GHI rRMSE |
GHI rMBE |
DNI rMAE |
DNI rRMSE |
DNI rMBE |
DHI rMAE |
DHI rRMSE |
DHI rMBE |
|
Mohe |
47.06 |
94.45 |
27.4 |
257.49 |
710.6 |
205.78 |
56.42 |
74.94 |
-21.16 |
|
Harbin |
43.29 |
90.06 |
24.53 |
186.88 |
656.5 |
122.55 |
45.8 |
58.7 |
-3.84 |
|
Urumqi |
53.3 |
66.43 |
29.7 |
101.76 |
146.4 |
71.19 |
41.86 |
55.28 |
-19.99 |
|
Kashi |
63.31 |
79.81 |
23.55 |
150.46 |
209.17 |
76.66 |
49.97 |
59.14 |
-14.54 |
|
Ejinaqi |
143.72 |
577.3 |
124.1 |
314.3 |
1358.6 |
285.07 |
62.15 |
77.02 |
-29.21 |
|
Geermu |
55.25 |
183.81 |
29.29 |
543.85 |
1141.35 |
423.95 |
206.12 |
257.3 |
-149.61 |
|
Yuzhong |
67.36 |
103.16 |
43.01 |
210.57 |
419.9 |
181.58 |
53.54 |
65.22 |
-27.36 |
|
Shenyang |
42.42 |
59.82 |
17.53 |
142.76 |
241.94 |
52.43 |
61.1 |
79.54 |
-1.34 |
|
Beijing |
41.91 |
75.55 |
14.51 |
149.81 |
484.88 |
72.96 |
50.35 |
65.64 |
-15.72 |
|
Lhasa |
66.12 |
95.98 |
40.92 |
165.98 |
279.67 |
128.37 |
58.18 |
67.78 |
-36.76 |
|
Chengdu |
300.02 |
1786.89 |
282.23 |
1630.53 |
11590.15 |
1611.59 |
72.87 |
90.41 |
19.66 |
|
Kunming |
57.46 |
72.59 |
19.9 |
124.82 |
193.57 |
79.38 |
52.83 |
65.11 |
-18.03 |
|
Zhengzhou |
49.47 |
63.42 |
27.52 |
200.62 |
300.57 |
151.38 |
44.5 |
57.09 |
-29.81 |
|
Wuhan |
57.29 |
73.02 |
28.46 |
256.05 |
402.22 |
194.52 |
49.75 |
62.58 |
-22.17 |
|
Baoshan |
55.33 |
109.96 |
35.16 |
278.34 |
898.74 |
217.92 |
49.24 |
64.44 |
-13.33 |
|
Guangzhou |
61.94 |
79.27 |
30.16 |
222.49 |
343.88 |
152.63 |
45.09 |
57.6 |
-16.56 |
|
Sanya |
53.76 |
67.65 |
19.54 |
158.62 |
220.96 |
83.88 |
44.96 |
56.16 |
-18.31 |
|
Xianghe |
59.75 |
202.47 |
33.63 |
2630.21 |
70224.38 |
2386.57 |
105.62 |
143.48 |
-65.1 |
Tab. 2: Performance statistics at Xianghe (unit: percent)
|
Method |
GHI rMAE |
GHI rRMSE |
GHI rMBE |
DNI rMAE |
DNI rRMSE |
DNI rMBE |
DHI rMAE |
DHI rRMSE |
DHI rMBE |
|
BRL |
59.75 |
202.47 |
33.63 |
2630.21 |
70224.38 |
2386.57 |
105.62 |
143.48 |
-65.1 |
|
Tuned BRL |
59.75 |
202.47 |
33.63 |
2584.57 |
70203.9 |
2336.69 |
99.86 |
134.63 |
-36.4 |
|
ANN |
30.87 |
40.64 |
-15.85 |
76.7 |
97.88 |
-30.54 |
37.47 |
51.23 |
6.89 |
Tab. 3: Performance statistics at Geermu (unit: percent)
|
Method |
GHI rMAE |
GHI rRMSE |
GHI rMBE |
DNI rMAE |
DNI rRMSE |
DNI rMBE |
DHI rMAE |
DHI rRMSE |
DHI rMBE |
|
BRL |
55.25 |
183.81 |
29.29 |
543.85 |
1141.35 |
423.95 |
206.12 |
257.3 |
-149.61 |
|
Tuned BRL |
55.25 |
183.81 |
29.29 |
534.18 |
1134.28 |
390.86 |
208.07 |
261.76 |
-119.08 |
|
ANN |
82.92 |
99.85 |
-82.38 |
97.85 |
141.98 |
-96.04 |
70.11 |
91.56 |
-67.05 |