Wind turbine power curve estimation based on earth mover distance and artificial neural networks

Abstract

A data-based estimation of the wind-power curve in wind turbines may be a challenging task due to the presence of anomalous data, possibly due to wrong sensor reads, operation halts, malfunctions or other. In this study, the authors describe a data-based procedure to build a robust and accurate estimate of the wind-power curve. In particular, they combine a joint clustering procedure, where both the wind speeds and the power data are clustered, with an Earth Mover Distance-based Extreme Learning Machine algorithm to filter out data that poorly contribute to explain the unknown curve. After estimating the cut-in and the rated speed, they use a radial basis function neural network to fit the filtered data and obtain the curve estimate. They extensively compared the proposed procedure against other conventional methodologies over measured data of nine turbines, to assess and discuss its performance.

Fault diagnosis of photovoltaic strings by using machine learning‐based stacking classifier

“…Here, W(l) ij is the connection weight between the j th neuron in layer l −1 and the i th neuron in layer l ; b(l) i is the bias of the i th neuron in layer l ; net(l) I is the input of the i th neuron in layer l ; f(x) is the activation function of the neurons. Therefore, the computations of the network can be expressed as follows [36, 37]: $$$$\begin{equation}net_i^{(l)} = \sum_{j = 1}^{{s}_l - 1} {W_{ij}^{(l)}} h_j^{(l - 1)} + b_i^{(l)}\end{equation}$$$$ $$$$\begin{equation}h_i^{(l)} = f\left( {{\rm{ net }}_i^{(l)}} \right)\end{equation}$$$$…”

“…Here, W(l) ij is the connection weight between the jth neuron in layer l−1 and the ith neuron in layer l; b(l) i is the bias of the ith neuron in layer l; net(l) I is the input of the ith neuron in layer l; f(x) is the activation function of the neurons. Therefore, the computations of the network can be expressed as follows [36,37]:…”

Hourly Water Level Forecasting in an Hydroelectric Basin Using Spatial Interpolation and Artificial Intelligence

“…Other methods, such as boosted decision trees and Bayesian linear regression were reported to be successful for daily water level prediction in a hydroelectric basin using also rainfall data as input [13]. Successful applications of machine learning approaches for time series forecasting in the energy sector can be found also in the case of electricity price prediction [19], where Kalman filer and Echo State Networks are used, electrical load prediction [20], where a variant of the K-Nearest Neighbours algorithm is proposed, as well as prediction of the power generation from photovoltaic plants [21], and wind plants [22,23].…”

Hourly Water Level Forecasting in an Hydroelectric Basin Using Spatial Interpolation and Artificial Intelligence

“…Other methods, such as boosted decision trees and Bayesian linear regression were reported to be successful for daily water level prediction in a hydroelectric basin using also rainfall data as an input [ 14 ]. Successful applications of machine learning approaches for time series forecasting in the energy sector can be found also in the case of electricity price prediction [ 20 ], where Kalman filter and Echo State Networks are used, electrical load prediction [ 21 ], where a variant of the K-Nearest Neighbours algorithm is proposed, as well as prediction of the power generation from photovoltaic plants [ 22 ], and wind plants [ 23 ]. In [ 24 ], the Kalman filter is successfully applied as a data assimilation scheme for water level forecasting.…”