Moving window kernel PCA for adaptive monitoring of nonlinear processes

Adaptive Weighting Just-in-Time-Learning Quality Prediction Model for an Industrial Blast Furnace

“…(5)), a weighted distance is adopted here: (12) where v k is the feature weight parameter to balance the components of z i , and z i,k indicates the k th element in z i . Inspired by the principle component analysis, 45) an eigenvalue based weighting strategy is formulated to calculate the importance of the latent variables: With Eq. (13), the element with a larger eigenvalue will be assigned a relative smaller importance (≈0) and can be ignored to reduce the projection dimension automatically.…”

Wide area phase angle measurements for islanding detection — An adaptive nonlinear approach

“…The construction of a moving window KPCA model involves the selection of a kernel parameter, σ, the window length, N, the initial number of principal components, r, and the delay for applying the adaptive KPCA model, l [23].…”

“…As the window slides along the data, a new nonlinear model is built by including the newest sample and removing the oldest one. While a number of algorithmic developments have been reported to update and downdate the models efficiently [22], [23], their computational accuracy may be compromised. To present the basic concept of adaptive KPCA, which is promising for islanding protection, the traditional moving window approach is used in this paper.…”

Model-based virtual sensors by means of Modelica and FMI

“…A review of data driven virtual sensors can be found in (Kadlec et al, 2011). Some common approaches include developing autoregressive models of the system as in (Samara et al, 2013), using artificial neural networks ( (Bizon et al, 2014), (Gonzaga et al, 2009)) or using moving window methods as in (Liu et al, 2009). The required data training may be a handicap in systems where data cannot be acquired continuously or in which faulty conditions cannot be measured.…”