application to electrical data
play

Application to Electrical Data Jean-Michel Poggi Laboratoire de - PowerPoint PPT Presentation

19th International Conference on Computational Statistics Paris - France, August 22-27 Empirical Mode Decomposition for Trend Extraction. Application to Electrical Data Jean-Michel Poggi Laboratoire de Mathmatique, Universit dOrsay,


  1. 19th International Conference on Computational Statistics Paris - France, August 22-27 Empirical Mode Decomposition for Trend Extraction. Application to Electrical Data Jean-Michel Poggi Laboratoire de Mathématique, Université d’Orsay, France Joint work with Farouk Mhamdi, Mériem Jaïdane-Saïdane U2S, ENIT, Tunis, Tunisia

  2. Outline • Motivation: Blind Trend Extraction • Empirical Mode Decomposition (EMD) • EMD trend vs. Hodrick Prescott (HP) trend • Simulated seasonal series • Tunisian daily peak load 2000-2006 • EMD vs. HP, wavelet trends This work was supported by 2005/2009 VRR research project between ENIT and Tunisian Society of Electricity and Gas (STEG) EMD for trend extraction 2

  3. Trend? • Trend = some “ smooth additive component that contain information about global change” Alexandrov et al. (2009) • The problem: to extract trends from seasonal time series without strong modeling and global estimation • Ex: long term electricity load or airline traffic forecasting EMD for trend extraction 3

  4. Seasonal time series intrinsic trend • Additive time series model : Y(t) = T(t) + SC(t) + e(t) i.e. trend + seasonal-cycles + error • Several methods are used for time series components extraction including – local or global regressions – MA filtering, X11, X12 – Hodrick Prescott filter see Alexandrov et al. (2009) for a recent review • The idea: to directly extract trend without identifying the other components of the observed signal EMD for trend extraction 4

  5. EMD? EMD and trend? EMD transformation is nonlinear and suitable for non-  stationary signals Huang et al. (2004) Signal is decomposed as a superposition of local sums of  oscillatory components called Intrinsic Mode Functions (IMF) of different time scales intrinsic to the signal Y(t) =  IMF k (t) + r(t) with IMF = function with zero mean and having as many zero crossings as maxima or minima and r = a monotone function IMFs are fully data-driven and local in time  After the IMF extraction process (sifting) , it remains  r(t) a monotone function candidate to be an estimate of T(t) since it is a trend free of oscillatory components EMD for trend extraction 5

  6. EMD extraction algorithm Residue = s(t), I 1 (t) = Residue IMF? i = 1, k = 1 while Residue not equal zero or not monotone while I i has non-negligible local mean U(t) = spline through local maxima of I i L(t) = spline through local minima of I i Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t), i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k Credit: Suz Tolwinski University of Arizona k = k+1 Program in Applied Mathematics Spring 2007 RTG EMD for trend extraction 6 end

  7. EMD acting on an example Analyzed signal = tone + chirp tone + chirp = tone + chirp Credit: Rilling and Flandrin from http://perso.ens-lyon.fr/patrick.flandrin/emd.html EMD for trend extraction 7

  8. EMD acting on an example iteration 0 - Analyzed signal = starting point IMF 1; iteration 0 Residue = s(t) 2 I 1 (t) = Residue 1 i = 1 0 k = 1 while Residue not equal zero or not monotone -1 while I i has non-negligible local mean -2 10 20 30 40 50 60 70 80 90 100 110 120 U(t) = spline through local maxima of I i L(t) = spline through local minima of I i Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 8

  9. EMD acting on an example Extract local maxima IMF 1; iteration 0 Residue = s(t) 2 I 1 (t) = Residue 1 i = 1 0 k = 1 while Residue not equal zero or not monotone -1 while I i has non-negligible local mean -2 10 20 30 40 50 60 70 80 90 100 110 120 U(t) = spline through local maxima of I i L(t) = spline through local minima of I i Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 9

  10. EMD acting on an example Maxima envelope by interpolation IMF 1; iteration 0 Residue = s(t) 2 I 1 (t) = Residue 1 i = 1 k = 1 0 while Residue not equal zero or not monotone -1 while I i has non-negligible local mean -2 U(t) = spline through local maxima of I i 10 20 30 40 50 60 70 80 90 100 110 120 L(t) = spline through local minima of I i Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 10

  11. EMD acting on an example Extract local minima Residue = s(t) IMF 1; iteration 0 I 1 (t) = Residue 2 i = 1 1 k = 1 0 while Residue not equal zero or not monotone while I i has non-negligible local mean -1 U(t) = spline through local maxima of I i -2 L(t) = spline through local minima of I i 10 20 30 40 50 60 70 80 90 100 110 120 Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 11

  12. EMD acting on an example Minima envelope by interpolation IMF 1; iteration 0 Residue = s(t) 2 I 1 (t) = Residue 1 i = 1 k = 1 0 while Residue not equal zero or not monotone -1 while I i has non-negligible local mean -2 U(t) = spline through local maxima of I i 10 20 30 40 50 60 70 80 90 100 110 120 L(t) = spline through local minima of I i Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 12

  13. EMD acting on an example Mean of maxima and minima envelopes IMF 1; iteration 0 Residue = s(t) 2 I 1 (t) = Residue 1 i = 1 0 k = 1 -1 while Residue not equal zero or not monotone while I i has non-negligible local mean -2 10 20 30 40 50 60 70 80 90 100 110 120 U(t) = spline through local maxima of I i L(t) = spline through local minima of I i Local low frequency component Av(t) = 1/2 (U(t) + L(t)) I i (t) = I i (t) - Av(t) i = i + 1 end IMF k (t) = I i (t) Residue = Residue - IMF k k = k+1 end EMD for trend extraction 13

  14. EMD acting on an example Subtract mean envelope from signal Residue = s(t) IMF 1; iteration 0 I 1 (t) = Residue 2 i = 1 1 k = 1 0 while Residue not equal zero or not monotone while I i has non-negligible local mean -1 U(t) = spline through local maxima of I i -2 L(t) = spline through local minima of I i 10 20 30 40 50 60 70 80 90 100 110 120 Av(t) = 1/2 (U(t) + L(t)) residue I i (t) = I i (t) - Av(t) (“residue” -->) 1.5 i = i + 1 1 end 0.5 0 IMF k (t) = I i (t) -0.5 Residue = Residue - IMF k -1 k = k+1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 end EMD for trend extraction 14

  15. EMD acting on an example Is the residue a IMF? No Residue = s(t) I 1 (t) = Residue i = 1 IMF 1; iteration 0 2 k = 1 1 while Residue not equal zero or not monotone while I i has non-negligible local mean 0 U(t) = spline through local maxima of I i -1 L(t) = spline through local minima of I i -2 Av(t) = 1/2 (U(t) + L(t)) 10 20 30 40 50 60 70 80 90 100 110 120 I i (t) = I i (t) - Av(t) (“residue” -->) residue i = i + 1 1.5 end 1 IMF k (t) = I i (t) 0.5 Residue = Residue - IMF k 0 -0.5 k = k+1 -1 end -1.5 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 15

  16. EMD acting on an example No, so iterate the loop (sifting) Residue = s(t) IMF 1; iteration 1 I 1 (t) = Residue 1.5 i = 1 1 k = 1 0.5 0 while Residue not equal zero or not monotone -0.5 while I i has non-negligible local mean -1 U(t) = spline through local maxima of I i -1.5 L(t) = spline through local minima of I i 10 20 30 40 50 60 70 80 90 100 110 120 Av(t) = 1/2 (U(t) + L(t)) residue I i (t) = I i (t) - Av(t) 1.5 i = i + 1 1 end 0.5 0 IMF k (t) = I i (t) -0.5 Residue = Residue - IMF k -1 k = k+1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 end EMD for trend extraction 16

  17. EMD acting on an example IMF1 - iteration 1 - maxima IMF 1; iteration 1 1.5 1 0.5 0 -0.5 -1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 residue 1.5 1 0.5 0 -0.5 -1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 17

  18. EMD acting on an example IMF1 - iteration 1 - minima IMF 1; iteration 1 1.5 1 0.5 0 -0.5 -1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 residue 1.5 1 0.5 0 -0.5 -1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 18

  19. EMD acting on an example IMF1 - iteration 2 – mean IMF 1; iteration 2 1.5 1 0.5 0 -0.5 -1 -1.5 10 20 30 40 50 60 70 80 90 100 110 120 residue 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 19

  20. EMD acting on an example IMF1 - iteration 3 IMF 1; iteration 3 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 residue 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 20

  21. EMD acting on an example IMF1 - iteration 4 IMF 1; iteration 4 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 residue 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 21

  22. EMD acting on an example IMF1 - iteration 5 IMF 1; iteration 5 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 residue 1 0.5 0 -0.5 -1 10 20 30 40 50 60 70 80 90 100 110 120 EMD for trend extraction 22

Recommend


More recommend