Robust Fault Detection of Wind Energy Conversion Systems Based on Dynamic Neural Networks

Table 1

The ARS training algorithm.

Step	Operation
(1) Initialization phase:	𝜃0, 𝑛max, 𝜐0 and 𝐽min should be selected and 𝑛=1, 𝜃best=𝜃0;
(2) Variance selection phase:	𝑖=1, 𝑘=1, 𝜃𝑘=𝜃0; while (𝑖<5); while (𝑘≤100/𝑖); The trial point algorithm is executed; 𝑘=𝑘+1; end; 𝑖=𝑖+1, 𝑘=1, 𝜃𝑘=𝜃0; end
(3) Variance exploration phase:	𝑘=1, 𝜃𝑘=𝜃best, 𝑖=𝑖best; while (𝑘≤100); The trial point algorithm is executed; 𝑘=𝑘+1; end; if ((𝑛=𝑛max) or (𝐽(𝜃best)<𝐽min)); Break; else 𝜃0=𝜃best, 𝑛=𝑛+1 and go to Step  2.; end
The trial point algorithm:	𝜐𝑖=10^−𝑖𝜐0, 𝜃𝑘^′=𝜃𝑘+𝑟𝑘; if (𝐽(𝜃𝑘^′)≤𝐽(𝜃𝑘)); 𝜃𝑘+1=𝜃𝑘^′; else 𝜃𝑘+1=𝜃𝑘; end; if (𝐽(𝜃𝑘^′)≤𝐽(𝜃best)); 𝜃best=𝜃𝑘^′; 𝑖best=𝑖; end