Download presentation
Presentation is loading. Please wait.
Published byAshlynn Ramsey Modified over 9 years ago
1
1 Neural Network Data Fusion and Uncertainty Analysis for Wind Speed Measurement using Ultrasonic Transducer Juan M. Mauricio Villanueva jmauricio12@gmail.com January, 2011
2
2 Introduction There is the need for the determination of the wind power density (WPD), which is used in eolic energy as requirements on wind turbine localization. where: is the air density and is the wind speed
3
3 Introduction The objective of the measurement procedure is to defined a criteria to ensure that the data: Sufficient quantity To determine the power and quality performance characteristic of the wind turbine accurately
4
4 Introduction The wind speed measurement should be supplemented with an estimate of the uncertainty of the measurement The uncertainty estimate is based on the ISO guide: “Guide to the expression of uncertainty in measurement”
5
5 Objetives The purpose of this paper are: Provide a procedure that will ensure consistency, accuracy and reproducibility into the wind speed measurement A data fusion procedure based on neural network algorithm to determine the fusion ToF Assessment the fusion uncertainty of a conventional ultrasonic transducer configuration
6
6 Wind Speed Measurement Transducers Configuration
7
7 Measurement Model and Data Fusion Procedures
8
8 The model is linear in the sense that the model output is a linear combination of its inputs.
9
9 Measurement and Uncertainty of ToF Analysis and assessment of uncertainty for ToF measurement through the TH and PD techniques are carried out. The ToF measurement by TH techniques and m=10 ToF measurement by PD techniques
10
10 Measurement and Uncertainty of ToF Uncertainty in measurement is a parameter associated with the result of a measurement. Following the ISO Guide, the uncertainties are expressed as standard deviations and are denoted standard uncertainties: where: uTh and uPD are the standard deviation values of the TH and PD techniques and uFusion is the standard deviation value of fusion
11
11 Results and Simulations We apply the data fusion procedure for the estimation of the ToF, combining independent information of the ToF obtained by the methods of TH and PD From these results, we can determine the measurements and their associated uncertainties
12
12 Results and Simulations The model is simulated in Simulink (MATLAB) Wind speed from 5 to 30 m/s One ToF estimation measurement by TH m=10 ToF estimation measurement by PD Transducers operating frequency: f = 40 kHz; Maximum voltage level: vm = 1 volt; Attenuation medium: Att = 10 % of vm; Additive uncertainty: uA equal to 0.01 volt; Frequency clock: fs = 50 MHz. uTH = 0.5 µs uPD = 0.1 µs
13
13 Results and Simulations ToF simulation values and uncertainties (in us) (m/s) ToF Theory (µs)ToF fusion (µs)u fusion (µs) 5239.57239.590.115 10237.86237.850.133 15236.16236.140.137 20234.49234.480.142 25232.85232.830.049 30231.22230.450.052
14
14 Results and Simulations From this results, we can make a Gaussian Distribution of ToF measurement fusion. For example, to the wind speed measurement 10 m/s:
15
15 Results and Simulations Gaussian Distribution of ToF measurement fusion.
16
16 Conclusions This paper presents a method to wind speed measurement based on neural network for multisensor fusion. Quantitatively, the fusion procedures increase the accuracy of inference, i.e. reduce the uncertainties in the ToF estimation. Qualitatively, the neural network fusion procedure take the advantages of the TH and PD techniques when used individually. The fusion algorithm produces a ToF results with less uncertainty, reducing ambiguity and increasing the reliability of measurement and, consequently, improving the operational performance of the measurement model.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.