Results
Long-term Predict
Result for four algorithms
For long-term prediction, ANN has the highest degree of fit and RNN has the smallest error. It is obvious from the figure that the linear regression model is not suitable for long-term prediction of solar energy data, which can be verified via large MAE and MSE. Regression Tree performs much better than linear regression in the long-term prediction but is not as good as the other two models. Both the MAE and MSE are larger than those of the ANN model and the RNN model.
Short-term Predict
Result for four algorithms
For the short-term prediction, each model fits the input data well, but in terms of error, the RNN model takes an absolute lead. Among the four algorithms, the error of the regression tree is extremely large, with the MSE greater than 8, and the MAE greater than 2. On the contrary, the RNN model has almost no error. It is believed that RNN has best performance in the short-term prediction. The ANN model’s prediction ranks second in terms of short-time prediction, with MAE and MSE less than 1.