tailieunhanh - AUTOMATION & CONTROL - Theory and Practice Part 8

Tham khảo tài liệu 'automation & control - theory and practice part 8', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | 166 AUTOMATION CONTROL - Theory and Practice Fig. 5. Generic structure for three layer ANN. The procedure to calculate the propagation values for each layer is as follows. The input layer receives a value Xi which is propagated with a weight Wij to the intermediate layer according to equation 3 where S is called the activation coefficient. Sjr ij w 1 3 This activation coefficient Sjp is propagated by an output function which represents the intermediate layer in this case the sigmoid function equation 4 . Ã S -1 4 1 e Propagation may be accomplished using alternative range-limited functions such as tanh 1 - selection of an appropriate function depends on the final application. As shown in figure 5 it is also necessary to define an exciting function in order to access the output layer equation 5 and this is obtained when the weights Vjk have been calculated. l n z w V 1 5 j 1 Finally the output function of the output neuron k is obtained using the same sigmoid function that was used to connect the input and hidden layers equation 6 . Okp 1 1 e-rk 6 Neural PDF The principal reason that Artificial Neural Networks ANNs have earned their position in the analysis design and implementation of control strategies is their flexibility. If it is possible to acquire any a priori knowledge of the structure of the model of a system to be controlled this information can be used to improve the tuning of typical controllers such as PD or PID and in the case considered here PDF. There are many contributions in the area of Neural PDF Control Strategy for a Hydroelectric Station Simulator 167 artificial neural networks aimed at defining fast and effective strategies to calculate and adjust the parameters of discrete PID control systems Narendra Mukhopadhyay 1996 Garcez Garcez 1995 . In this work a similar strategy is used to tune a discrete PDF. There have been several works where ANN have been applied to hydroelectric systems. Garcez and Garcez 1995 applied PI neural control to a