Table 3 Performance of PV power generation prediction models under different meteorological conditions.

From: A novel hybrid model integrating CEEMDAN decomposition, dispersion entropy and LSTM for photovoltaic power forecasting and anomaly detection

Weather

Model

RMSE

MAE

SSE

nRMSE

nMAE

\(\:{\mathbf{R}}^{2}\)

Sunny

GRU

0.406

0.367

47.471

0.069

0.062

0.966

LSTM

0.337

0.288

32.672

0.058

0.049

0.977

Transformer

0.239

0.161

16.384

0.041

0.027

0.988

DLinear

0.230

0.143

15.095

0.039

0.024

0.989

CEEMDAN-LSTM

0.220

0.141

13.832

0.037

0.024

0.987

CEEMDAN-DispEn-LSTM

0.198

0.109

11.321

0.034

0.019

0.991

Cloudy

GRU

0.731

0.698

153.700

0.158

0.151

0.116

LSTM

0.599

0.551

103.290

0.130

0.120

0.406

Transformer

0.392

0.291

44.295

0.085

0.063

0.745

DLinear

0.355

0.200

36.364

0.077

0.043

0.800

CEEMDAN-LSTM

0.340

0.267

33.300

0.073

0.036

0.780

CEEMDAN-DispEn-LSTM

0.328

0.138

30.981

0.071

0.029

0.806

Light rain turning to cloudy

GRU

0.462

0.425

61.517

0.128

0.117

0.612

LSTM

0.416

0.345

49.863

0.115

0.095

0.686

Transformer

0.277

0.160

22.141

0.077

0.044

0.860

DLinear

0.293

0.175

24.795

0.081

0.048

0.844

CEEMDAN-LSTM

0.274

0.145

21.676

0.076

0.040

0.863

CEEMDAN-DispEn-LSTM

0.264

0.115

20.088

0.073

0.031

0.873

Moderate rain

GRU

0.435

0.417

54.520

0.262

0.251

-1.483

LSTM

0.274

0.248

21.571

0.165

0.149

0.018

Transformer

0.153

0.137

6.711

0.092

0.083

0.694

DLinear

0.173

0.150

8.639

0.105

0.088

0.607

CEEMDAN-LSTM

0.117

0.080

3.904

0.070

0.048

0.822

CEEMDAN-DispEn-LSTM

0.100

0.050

2.879

0.070

0.030

0.869

  1. The bold values are the optimal values among all methods.