Performance Analysis of a Wind Turbine Pitch Neurocontroller with Unsupervised Learning

Algorithm 1

Proposed learning algorithm.

	% Initialization
	Xmin ⟵ PerrMin
	Xmax ⟵ PerrMax
	Ymin ⟵ dotPerrMin
	Ymax ⟵ dotPerrMax
	IncX ⟵ (Xmax − Xmin)/(Nx − 1)
	IncY ⟵ (Ymax − Ymin)/(Ny − 1)
	M ⟵ Nx  Ny
	for i = 0 toM − 1
	cNetX ⟵ (i DIV Ny)  IncX + XMin
	cNetY ⟵ (i MOD Ny)  IncY + YMin
	cNet (i) ⟵ (cNetX, cNetY)
	W (i) ⟵ 1
	Fold (i) ⟵ 0
	end for
	F ⟵ Fold
	tconOld ⟵ 0
	pitchCon ⟵ 0
	% Execute algorithm
	(errPow, derrPow, errPowSum) ⟵ MODEL (0)
	for t = 0 to tEnd
	if t ≥ tconOld + Tc then
	errPowSat ⟵ MIN (Xmax, MAX (Xmin, errPow))
	derrPowSat = MIN (Ymax, MAX (Ymin, derrPow))
	[RBFout, F] = RBF(cNet, W, errPowSat, derrPowSat)
	Fold ⟵ F
	pitchCon ⟵ (pi/4) − RBFout
	if ABS (errPowSat) < minErr then
	Winc ⟵ 0
	else
	errM ⟵ errPowSat  KP + derrPowSat  KD + errPowSum  KI
	Winc ⟵ Fold  errM  mu
	end if
	W ⟵ W + Winc
	tconOld ⟵ t
	end if
	(errPow, derrPow, errPowSum) ⟵ MODEL (pitchCon)
	end for