4. Forecasting the electricity generation

From the error histogram (Figure 7), it can be observed that the errors are between �0.13 and

We trained the network using the three algorithms (LM, RB and SCG), the best results being recorded using the Bayesian Regularization algorithm, although the Levenberg-Marquardt

In Table 3, the results obtained with autoregressive neural networks are compared with

The accuracy of ANN algorithms is better (about 95%) compared to the accuracy of stochastic models. Also, the Levenberg-Marquardt and Bayesian regularization algorithms are also superior regarding the lowest MSE. The R coefficient and error distribution for neural network

LM RB GCS AR MA ARMA ARIMA

�0.3 to 0.12 �0.13 to 0.12 �0.18 to 0.22 �1.24 to 1.16 �1.36 to 1.44 �1.11 to 0.99 �1.14 to 0.66

MSE 0.0064 0.0046 0.167 0.0091 0.0275 0.0316 0.0287 MAPE 4.26 4.21 6.21 7.29 24.45 29.05 24.97

Table 3. Autoregressive neural networks versus stochastic methods.

algorithm recorded good results with an increased performance in training.

algorithms are better than AR, MA, ARMA and ARIMA models.

+0.12, which can be considered an acceptable distribution.

stochastic methods (ARMA, ARIMA and AR).

130 Advanced Applications for Artificial Neural Networks

Figure 7. Errors histogram.

Performance/ method

Errors distribution In this section we will analyze stochastic methods based on ARMA and ARIMA models compared with feed-forward artificial neural networks for small wind turbines and photovoltaic panels generation in case of short-term forecasting.
