| References | Methods | TH | Inputs | Best results |
| [18] | Persistence, MPL, CNN, LSTM, and LSTM full. | VST | Past PV power data and sky images. | RMSE = 15.3%. |
| [19] | The component methods including SARIMA, ETS, MLP, STL, TBATS, theta, NWP, MOS, temporal reconciliation (TmpRec), and geographical reconciliation (GeoRec). The combined forecasts including simple averaging, Var, ordinary least squares (OLS), least absolute deviation (LAD), constrained least squares (CLS), subset, AIC, lasso, and Oracle. | ST | Past PV power. | NRMSE = 15.4%. |
| [20] | Probabilistic forecast based on the Gaussian process (GP) and the reference model based on ARIMA. | 0–6 h | Household electricity consumption and past PV power. | NRMSE = 8.2% PINAW: 12.4% PICP: 87.57%. |
| [21] | GA + PSO + ANFIS compared to BPNN, and LRM. | ST | Past PV power data and NWP data. | NRMSE = 5.48%. |
| [22] | WT, FNN, ELM, and cascade forward BPNN (NewCF) learned with different learning methods. | ST | Past PV power, air temperature, wind speed, and humidity. | MAPE = 3.10%. |
| [23] | RF, fuzzy C-means (FCM), sparse Gaussian process (SPGP), and improved grey wolf optimizer (IMGWO). | ST | Past PV power data. | NRMSE = 6.5%. |
| [24] | Models for clear sky weather: SARIMA, W-SARIMA, RVFL, W-RVFL, and SVR. Models for cloudy/rainy weather: SARIMA-RVFL hybrid model. | VST | PV power data. | RMSE = 9.34%. |
| [25] | SVM, MLP, multivariate adaptive regression spline (MARS), and SVM-MLP-MARS. | ST | Past PV power, wind speed, wind direction, temperature, relative humidity, GHI, and DHI. | RMSE = 21.41%. |
| [26] | CNN | VST | PV power and sky images. | RMSE = 2.5 kW. |
| [27] | CNN, LSTM, and the hybrid model of CNN-LSTM. | ST | Wind speed, temperature, relative humidity, GHI, DHI, wind direction, current phase average, and active power. | RMSE = 0.9 kW. |
| [7] | Uncertain basis function method (UBF): UBU (uniform), UBG (Gaussian), and UBP (Laplace). Stochastic state-space method (STS): prediction minimization error and expectation maximization and Kalman filter (EM-KF). | VST | Past PV power and solar irradiance. | NRMSE = 8.11% MAPE = 5.81 %. |
| [28] | CNN with the rectified linear activation function (RLAF), the multiheaded CNN of 4 CNNs, the CNN-LSTM, and the ARMA. | ST | PV power, irradiation, module and ambient temperatures, and wind speed. | RMSE = 0.046 kW. |
| [29] | CNN, residual network (RN), dense convolutional network (DCNN), theta, ETS, SVR, RFR, physical, MPL, and the hybrid of RN-DCNN. | ST | Past PV power and NWP data. | MSE = 0.152 kW. |
| [30] | Hoff, Perez, Lave, variability reduction index (VRI)—gene expression programming (GEP) and WT-ANFIS models. | 0–6 h | Irradiance data and weather conditions. | RMSE = 9.52 %. |
| [31] | Similarity algorithm (SA), KNN, NARX, and smart persistence models (SPMs). | ST | Past PV power, air and module temperatures, wind speed, wind direction, humidity, and solar irradiance. | RMSE = 2.3% RMSE = 0% RMSE = 5.9%. |
|
|