Embedded Implementation of Power System Monitoring Algorithms

Raymond McNamara, 09505075Electrical Energy Systems

FYP Presentation, January 2013

Introduction• Develop & implement numerous algorithms in

real-time for monitoring and control of power systems.

• Artificial signal generation on Matlab.• Compare filter-bank approaches with spectral

analysis approaches.(Performance and complexity)

• Port and evaluate the algorithm to a suitable real-time embedded platform.

• Develop & evaluate functionality for a suitable closed-loop control algorithm in Matlab.

• Port & evaluate closed-loop control algorithm to real-time embedded platform.

Research

• Looking at limits of class C equipment(Lighting equipment)• Accuracy of 1% replicating that of the ADE7880 Energy Meter

Reference: www.ieee.li

Harmonic Max. % of Current

n %A

2 2

3 30

5 10

7 7

9 5

11 n 39 3

Filter Bank Approach

Fast Fourier Transform method

Notch Filter

• Added to remove the peak at the first harmonic component with magnitude 1.

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04-100

-50

0

50

Normalized Frequency ( rad/sample)

Pha

se (

degr

ees)

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04-20

-10

0

10

Normalized Frequency ( rad/sample)

Mag

nitu

de (

dB)

Frequency Response of the Notch Filter

Transfer function for the filter:

>>freqz(Numerator Coefficients, Denominator Coefficients)

0.7 0.8 0.9 1 1.1 1.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

Real Part

Imag

inar

y P

art

Zplane Pole-Zero Diagram for the Notch Filter

0 500 1000 1500 20000

20

40

60

80

100

120

140

160

180

Number of samples

Mag

nitu

de o

f Y

(t)

FFT of signal after Notch filter and IIR filter

Second order system IIR filter(Resonator)

• Filters each harmonic separately.• Removes gain.• First 2000 samples removed due to filter

implementation. Transfer function for the filter:

0 100 200 300 400 500 6000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1Filter with gain removed

0 500 1000 1500 2000 2500 3000 3500

-50

0

50

100

150

200

250

300

350

400

450

Number of samples

Mag

nitu

de o

f Y(t)

FFT of signal after Notch filter and IIR filter

2000 4000 6000 8000 10000 12000 14000

0

100

200

300

400

500

600

Number of samples

Mag

nitu

de o

f Y

(t)

FFT of signal after Notch filter and IIR filter

2000 4000 6000 8000 10000 12000

0

100

200

300

400

500

600

700

Number of samples

Mag

nitu

de o

f Y

(t)

FFT of signal after Notch filter and IIR filter

50 Hz

1000 Hz 1950Hz

Gain removed

=1

Fast Fourier transform Method

• Zero-padding with next nearest power of 2 greater than the number of original samples ( 66536 instead of 51000).

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Frequency(Hz)

Mag

nitu

de

Magnitude Frequency Response after FFT

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Frequency(Hz)

Mag

nitu

de

Magnitude Frequency Response after FFT

Performance and Computational Complexity

• Assuming 5 seconds and 51000 samples. (5 x 51000) = 255,000.

• Notch & IIR Filter – 6 & 4 multiplies and 4 & 2 adds.(1 & 39 harmonics respectively) =1 (255,000x6)+39(255,000x4)mul & 1 (255,000x4) & 39(255,000x2)adds.

• Total = 41310000 + 20910000= 62,220,000.• FFT and inverse= 2(2Nlog2N) = 18,319,340• Multiplication : 4N = 1,020,000• Total = 19339340. • Saving of 68.9% with FFT

Future Plans

• Sort out Zero-padding within the FFT to make the algorithim more efficient with the DSP chip.

• Select a DSP chip that will have the capability of handling the data.

• Hopefully all going well, implement a closed control loop to monitor and adjust.

Conclusion

• For futher information about the project: http://harmonicalgorithm.wordpress.com/

• Thank you for your time and I hope you have enjoyed the presentation.

• Any Questions?