0. POWERLINE COMMUNICATIONS FOR ENABLING SMART GRID APPLICATIONS
Task ID:
Prof. Brian L. Evans
Wireless Networking and Communications Group
Cockrell School of Engineering
The University of Texas at Austin
May 3, 2013

1. Principal Investigator:
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Task Description:
Improve powerline communication (PLC) bit rates for monitoring/controlling applications for residential and commercial energy uses
Anticipated Results:
Adaptive methods and real-time prototypes to increase bit rates in PLC networks
Principal Investigator:
Prof. Brian L. Evans, The University of Texas at Austin
Current Students (with expected graduation dates):
Ms. Jing Lin Ph.D. (May 2014) Summer 2013 intern at TI
Mr. Yousof Mortazavi Ph.D. (Dec. 2013)
Mr. Marcel Nassar Ph.D. (Aug. 2013) Defended PhD April 15, 2013
Mr. Karl Nieman Ph.D. (May 2015) Summer 2013 intern at Freescale
Industrial Liaisons:
Dr. Anuj Batra (TI), Dr. Anand Dabak (TI), Mr. Leo Dehner (Freescale), Mr. Michael Dow (Freescale), Dr. Il Han Kim (TI), Mr. Frank Liu (IBM), Dr. Tarkesh Pande (TI) and Dr. Khurram Waheed (Freescale)
Starting Date: August 2010

2. Task Deliverables Date Tasks Dec 2010
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Task Deliverables
Date
Tasks
Dec 2010
Uncoordinated interference in narrowband PLC: measurements, modeling, and mitigation
May 2011
Testbed #1 based on TI PLC modems to investigate receiver improvements
Dec 2011
Narrowband PLC channel/noise: measurements/modeling
May 2012
Standard-compliant receiver methods (3x bit rate increase)
Dec 2012
Testbed #2 based on Freescale PLC modems to investigate transmitter improvements (2x bit rate increase)
On-going
Testbed #3 based on NI equipment to map noise mitigation algorithms onto FPGAs
Testbed #4 for two-transmitter two-receiver (2x2) systems based on TI PLC modems to investigate scalability

3. Recent Project Highlights
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Recent Project Highlights
Paper in Smart Grid Special Issue (Sep. 2012)
IEEE Signal Processing Magazine (impact factor 4.066)
Paper on channel impairments, noise, and standards
Co-authored with Dr. Anand Dabak (TI) and Dr. Il Han Kim (TI)
Channel Model Adopted (Oct. 2012)
Reference model for IEEE Standard for Low Frequency Narrow Band Power Line Communications for Smart Grid App.
Mr. Marcel Nassar, Dr. Anand Dabak (TI), Dr. Il Han Kim (TI), et al.
SRC Technical Transfer Talk (Dec. 2012)
Best Paper Award (Mar. 2013)
2013 IEEE Int. Symp. On Power Line Comm. and Its Applications
Co-authored with Dr. Khurram Waheed (Freescale)

4. Smart Grid Wind farm Central power plant HV-MV Transformer
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Smart Grid
Wind farm
Central power plant
HV-MV Transformer
Grid status monitoring
Utility control center
Smart meters
Integrating distributed energy resources
Houses
Offices
Traditional power grids have been coupled with communication networks today, leading to so-called smart grids. A smart grid enables information flow among various components of the grid, ranging from power plants to distributed energy resources, and from local utilities to residential and commercial customers. The purpose is to better monitor and control power generation and consumption. For example, distributed energy resources, such as solar power, fuel power and other personally-owned energy storage, can be integrated to the grid to provide low-cost or standby energy during peak hours. Transducers deployed over the grids are used to collect measurement data for grid status estimation and outage detection. Smart meters can be used for time-dependent pricing, which motivates customers to scale back their energy usage during peak hours, and also device-specific billing, so that the house owner doesn’t have be pay electric bills for charging his friend’s electric vehicle. In addition, smart buildings and smart homes can be energy efficient with automated lights, air conditioners and other smart appliances.
Device-specific billing
Automated control for smart appliances
Medium Voltage (MV) 1 kV – 33 kV
High Voltage (HV) 33 kV – 765 kV
Industrial plant

5. Smart Grid Goals Accommodate all generation types
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Smart Grid Goals
Accommodate all generation types
Renewable energy sources
Energy storage options
Improve operating efficiencies
Scale voltage with energy demand
Reduce peak demand
Analyze customer load profiles and system load snapshots
Improve system reliability
Power quality monitoring
Remote disconnect/reconnect
Outage/restoration event notification
Inform customer
ISTOCKPHOTO.COM/© SIGAL SUHLER MORAN
Enabled by smart meter communications
Source: Jerry Melcher, IEEE Smart Grid Short Course, 22 Oct. 2011, Austin TX USA

6. Smart Meter Communications
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Smart Meter Communications
Local utility
MV-LV transformer
Smart meters
Data concentrator
Communication backhaul
carries traffic between concentrator and utility on wired or wireless links
Smart meter communications
between smart meters and data concentrator via powerline or wireless links
On the distribution side of the grids, the communications between a local utility control center and its customers plays an important role in local utility applications that I just mentioned. The local utility communicates with a number of data concentrators located at the medium-voltage lines in the US via communication backhaul. Each data concentrator is in charge of a few houses and can talk to them via the last mile communications to the smart meters. The smart meter serves as a gateway in the home area sensor networks that interconnect transducers on appliances and indoor power lines.
Distances from Smart Meter to Data Concentrator: m in Europe/China/India and 3-4 km in US 3-4
Low voltage (LV) under 1 kV
Home area data networks
connect appliances, EV charger and smart meter via powerline or wireless links

7. Powerline Communications (PLC)
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Powerline Communications (PLC)
Use orthogonal frequency division multiplexing (OFDM)
Communication challenges
Channel distortion
Non-Gaussian noise
Categories
Band
Bit Rates
Coverage
Enables
Standards
Narrowband
3-500 kHz
~500 kbps
Multi-kilometer
Smart meter communication
(ITU) PRIME, G3
ITU-T G.hnem
IEEE P1901.2
Broadband
MHz
~200 Mbps
<1500 m
Home area data networks
HomePlug
ITU-T G.hn
IEEE P1901
The smart grid communications are supported by a heterogeneous set of network technologies, ranging from wireless to wireline solutions. Among the wireline alternatives, powerline communications, or PLC, have been deployed outdoor for last mile communications and indoor for home area networks. Narrowband PLC operating in the kHz band to deliver a few hundred kbps has been used for last mile communications over MV and LV lines. For home area networks, broadband PLC can provide several hundred Mbps in the MHz band. These PLC systems adopt multicarrier communications, or OFDM. (Introduce OFDM here)

8. OFDM Systems in Impulsive Noise
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
OFDM Systems in Impulsive Noise
FFT in receiver spreads impulsive energy over all tones
Signal-to-noise ratio (SNR) in each subchannel decreases
Narrowband PLC systems operate -5 dB to 5 dB in SNR
Data subchannels carry same number of bits (1-4) in current standards
Each 3 dB increase in SNR on data subchannels could give extra bit

9. Narrowband PLC Systems
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Narrowband PLC Systems
Problem: Non-Gaussian impulsive noise is #1 limitation to communication performance yet traditional communication system design assumes additive noise is Gaussian
Goal: Improve comm. performance in impulsive noise
Approach: Statistical modeling of impulsive noise
Solution #1: Receiver design (standard compliant)
Solution #2: Joint transmitter-receiver design
Parametric Methods
Nonparametric Methods
Listen to environment
No training necessary
Find model parameters
Learn statistical model from communication signal structure
Use model to mitigate noise
Exploit sparsity to mitigate noise

10. Narrowband PLC Impulsive Noise
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Narrowband PLC Impulsive Noise
Cyclostationary Noise
Asynchronous Noise
Example: rectified power supplies
Example: uncoordinated interference
Rx Receiver
Dominant in outdoor PLC
Increases with widespread deployment

11. Non-Parametric Mitigation Methods
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Non-Parametric Mitigation Methods
Exploit sparsity of impulsive noise in time domain
Build statistical model each OFDM symbol using sparse Bayesian learning (SBL)
At receiver, null tones contain only Gaussian + impulsive noise
SNR gain vs. conventional OFDM systems at symbol error rate 10-4
Complex, 128-point FFT, QPSK, data tones , rate ½ conv. code
Asynchronous Gaussian mixture model and Middleton Class A noise
time
System
Noise
SBL w/ null tones
SBL w/ all tones
SBL w/ decision feedback
Uncoded
GMM
8 dB
10 dB -
MCA
6 dB
7 dB
Coded
2 dB
9 dB
1.75 dB
6.75 dB
8.75 dB

12. Time Domain Interleaving
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Time Domain Interleaving
Bursts span consecutive OFDM symbols
Coded performance in cyclostationary noise
Interleave
Complex OFDM, 128-point FFT, QPSK, data tones , rate ½ conv. code
Bursts spread over many OFDM symbols

13. Time-Domain Interleaving
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Time-Domain Interleaving
Coded performance in cyclostationary noise
Burst duty cycle 10%
Burst duty cycle 30%
Time-domain interleaving over an AC cycle
Current PLC standards use frequency-domain interleaving (FDI)

14. Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Adaptive signal processing algorithms for bit loading and interference mitigation
Hardware
Software
NI x86 controllers stream data
NI cards generates/receives analog signals
TI front end couples to power line
Transceiver algorithms in C on x86
Desktop LabVIEW configures system and visualizes results
1x1 Testbed

15. Testbed #2: Noise Playback/Analysis
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Testbed #2: Noise Playback/Analysis
G3 link using two Freescale PLC modems
Freescale software tools allow frame-by-frame analysis
Test setup allows synchronous noise injection into power line
Freescale PLC G3-OFDM Modem
One modem to sample powerline noise in field
Collected 16k 16-bit 400 kS/s at each location
Freescale PLC Testbed

16. Testbed #2: Cyclic Power Line Noise
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Testbed #2: Cyclic Power Line Noise
Analyzed cyclic properties of PLC noise measurements
Developed cyclic bit loading method for transmitter
Receiver measures noise power over half AC cycle
Feedback modulation map to transmitter
Allocate more bits in higher SNR subchannels
2x increase in bit rate
Won Best Paper Award at ISPLC

17. Testbed #3: FPGA Implementation
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Testbed #3: FPGA Implementation
Built NI/LabVIEW testbed with real-time link (G3 settings)
Redesigned parametric impulsive noise mitigation algorithm
Converted matrix operations to distributed calculations on scalars
Based on approximate message passing (AMP) framework
Mapped transceiver to fixed-point data/math using Matlab
Synthesis: LabVIEW DSP Diagram to Xilinx Vertex 5 FPGAs
Received QPSK constellation at equalizer output
Utilization
Trans.
Rec.
AMP+Eq
FPGA 1 2 3
total slices
32.6%
64.0%
94.2%
slice reg.
15.8%
39.3%
59.0%
slice LUTs
17.6%
42.4%
71.4%
DSP48s
2.0%
7.3%
27.3%
blockRAMs
7.8%
18.4%
29.1%
conventional receiver
with AMP

18. Testbed #4: 2x2 PLC (On-Going)
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Testbed #4: 2x2 PLC (On-Going)
Goal: Improve communication performance by another 2x
One phase, neutral, ground for 2x2 differential signaling
Crosstalk between two channels due to energy coupling
Frequency response of a direct channel
Crosstalk highly correlated with direct channel response

19. Conclusion PLC systems are interference limited
Task Summary | Background | Noise Modeling and Mitigation | Testbeds | Conclusion
Conclusion
PLC systems are interference limited
Statistical models for interference
Cyclostationary model for impulsive noise synchronous to AC cycle
Gaussian mixture model for asynchronous impulsive noise
Interference management
Cyclic bit loading to double bit rates in cyclostationary noise
Time-domain interleaving to mitigate cyclostationary noise followed by receiver impulsive noise mitigation
Mapping impulsive noise mitigation algorithms to FPGAs
Poor: Non-parametric sparse Bayesian learning algorithms
Good: Parametric distributed approximate message algorithms

20. Our Publications Tutorial/Survey Article
M. Nassar, J. Lin, Y. Mortazavi, A. Dabak, I. H. Kim and B. L. Evans, “Local Utility Powerline Communications in the kHz Band: Channel Impairments, Noise, and Standards”, IEEE Signal Processing Magazine, Special Issue on Signal Processing Techniques for the Smart Grid, Sep Impact Factor
Journal Paper
J. Lin, M. Nassar and B. L. Evans, “Impulsive Noise Mitigation in Powerline Communications using Sparse Bayesian Learning”, IEEE Journal on Selected Areas in Communications, Special Issue on Smart Grid Communications, Jul Impact Factor
Conference Publications (more on next slide)
J. Lin and B. L. Evans, “Non-parametric Mitigation of Periodic Impulsive Noise in Narrowband Powerline Communications”, Proc. IEEE Global Communications Conference, Dec. 2013, Atlanta, GA USA, submitted.

21. Our Publications Conference Publications (more on next slide)
M. Nassar, P. Schniter and B. L. Evans, “Message-Passing OFDM Receivers for Impulsive Noise Channels”, Proc. Asilomar Conf. on Signals, Systems, and Computers, Nov. 2013, Pacific Grove, CA, submitted.
K. F. Nieman, M. Nassar, J. Lin and B. L. Evans, “FPGA Implementation of a Message-Passing OFDM Receiver for Impulsive Noise Channels”, Proc. Asilomar Conf. on Signals, Systems, and Computers, Nov. 2013, Pacific Grove, CA, submitted.
K. Nieman, J. Lin, M. Nassar, K. Waheed, and B. L. Evans, “Cyclic Spectral Analysis of Power Line Noise in the kHz Band”, Proc. IEEE Int. Sym. on Power Line Comm. and Its App., Mar. 2012, Johannesburg, South Africa. Best Paper Award.
J. Lin and B. L. Evans, “Cyclostationary Noise Mitigation in Narrowband Powerline Communications”, Proc. APSIPA Annual Summit and Conf., invited paper, Dec. 2012, Hollywood, CA USA.
M. Nassar, A. Dabak, I. H. Kim, T. Pande and B. L. Evans, “Cyclostationary Noise Modeling In Narrowband Powerline Communication For Smart Grid Applications”, Proc. IEEE Int. Conf. on Acoustics, Speech, and Signal Proc., Mar. 2012, Kyoto, Japan

22. Our Publications Conference Publications (more on next slide)
M. Nassar, K. Gulati, Y. Mortazavi, and B. L. Evans, “Statistical Modeling of Asynchronous Impulsive Noise in Powerline Communication Networks”, Proc. IEEE Int. Global Communications Conf., Dec. 2011, Houston, TX USA.
J. Lin, M. Nassar and B. L. Evans, “Non-Parametric Impulsive Noise Mitigation in OFDM Systems Using Sparse Bayesian Learning”, Proc. IEEE Int. Global Communications Conf., Dec. 2011, Houston, TX USA.
Standards Contribution
A. Dabak, B. Varadrajan, I. H. Kim, M. Nassar, and G. Gregg, “Appendix for noise channel modeling for IEEE P1901.2”, IEEE P Std., June 2011, doc: 2wg PHM5. Adopted as reference noise model in Oct ballot.

23. Thank you for your attention…
Questions?

24. Backup Slides

25. Today’s Power Grids in USA
7 large-scale power grids each managed by a regional utility company
700 GW generation capacity in total for long-haul high-voltage power transmission
Synchronized independently, and exchange power via DC transfer
130+ medium-scale power grids each managed by a local utility
Local power distribution to residential, commercial and industrial customers
Heavy penalties in US for blackouts (2003 legislation)
Utilities generate expected energy demand plus 12%
Energy demand correlated with time of day
Effect of plug-in electric vehicles (EVs) on energy demand uncertain
Generation cost 30x higher during peak times vs. normal load
Traditional ways to increase capacity to meet peak demand increase
Build new large-scale power generation plant at cost of $1-10B if permit issued
Build new transmission line at $0.6M/km which will take 5-10 years to complete
Source: Jerry Melcher, IEEE Smart Grid Short Course, 22 Oct. 2011, Austin TX USA

26. Comparison of Wireless & PLC Systems
Wireless Communications
Narrowband PLC (3-500 kHz)
Time selectivity
Time-selective fading and Doppler shift (cellular)
Periodic with period of half AC main freq. plus lognormal time-selective fading
Power loss vs. distance d
d –n/2 where n is propagation constant
e – a(f) d plus additional attenuation when passing through transformers
Propagation
Dynamically changing
Determinism from fixed grid topology
Synchronization
Varies
AC main power frequency
Additive noise/ interference
Assumed stationary and Gaussian
Gaussian plus non-Gaussian noise dominated by cyclostationary component
Asynchronous interference
Uncoordinated users in Wi-Fi bands; Frequency reuse in cellular
Due to power electronics and uncoordinated users using other standards
MIMO
Standardized for Wi-Fi and cellular
Number of wires minus 1; G.9964 standard for broadband PLC

27. Cyclostationary Noise Asynchronous Impulsive Noise
PLC Noise Scenarios
Background Noise
Cyclostationary Noise
Asynchronous Impulsive Noise
Spectrally shaped noise
Decreases with frequency
Superposition of lower-intensity sources
Includes narrowband interference
Cylostationary in time and frequency
Synchronous and asynchronous to AC main frequency
Comes from rectified and switched power supplies (synchronous), and electrical motors (asynchronous)
Dominant in narrowband PLC
Impulse duration from micro to millisecond
Random inter-arrival time
50dB above background noise
Caused by switching transients and uncoordinated interference
Present in narrowband and broadband PLC
time

28. Cyclostationary Noise
Noise Sources
Noise Trace

29. Uncoordinated Interference Results
Homogeneous PLC Network
General PLC Network

30. Cyclostationary Noise Modeling
Measurement data from UT/TI field trial
Cyclostationary Gaussian Model [Katayama06]
Proposed model uses three filters [Nassar12]
Demux
Period is one half of an AC cycle
s[k] is zero-mean Gaussian noise
Adopted by IEEE P narrowband PLC standard

31. Asynchronous Noise Modeling
Dominant Interference Source
Ex. Rural areas, industrial areas w/ heavy machinery
Middleton Class A Distribution [Nassar11]
Impulse rate l Impulse duration m
Homogeneous PLC Network
li = l, mi = m, g(di) = g0
Ex. Semi-urban areas, apartment complexes
Middleton Class A Distribution [Nassar11]
General PLC Network
li, mi, g(di) = gi
Ex. Dense urban and commercial settings
Gaussian Mixture Model [Nassar11]
Middleton Class A is a special case of the Gaussian Mixture Model.

32. Parametric vs. Nonparametric Methods
Must build a statistical model of the noise
Yes No
Requires training data to compute model parameters
Degrades in performance due to model mismatch
Has high complexity when receiving message data
Prior receiver methods to cope with non-Gaussian noise can be categorized as parametric and nonparametric approaches. Parametric methods assumes a statistical noise model, estimates model parameters during a training stage, and use that to mitigate noise during data transmissions. These methods generally allow low-complexity implementations during the transmission stage. However, in time-varying noise statistics, re-training is needed which introduces extra overhead. Otherwise performance degradation can be expected due to model mismatch. Nonparametric methods, on the other hand, require no additional training since they don’t assume any noise models. Instead they denoise the received signal by exploiting certain sparsity structure of the noise. These methods are more robust in various noise environments since it doesn’t rely on statistical models but the sparse structure of the noise. However, the denoising algorithms are generally of high computational complexity.

33. Asynchronous Noise Sparse in time domain Learn statistical model
Use sparse Bayesian learning (SBL)
Exploit sparsity in time domain [Lin11]
SNR gain of 6-10 dB
Increases 2-3 bits per tone for same error rate - OR -
Decreases bit error rate by x for same SNR
time
~10dB
~6dB
Transmission places 0-3 bits at each tone (frequency). At receiver, null tone carries 0 bits and only contains impulsive noise.

34. Performance w/o Error Correction
NSI
Gaussian mixture model noise
Non-parametric methods in blue
Parametric methods in red
Proposed
CS+LS: [Caire08] MMSE: [Haring02] SBL: [Lin11]

35. Performance w/ Error Correction
NSI
Proposed
Non-parametric methods in blue
Parametric methods in red
NSI
Gaussian mixture model noise

36. Power Line Noise at Residential Site
frequency sweep f = 170 kHz
narrowband f = 140 kHz
complex spectrum f = kHz

37. Analysis of Residential Noise
though spectrally complex, many components have strong stationarity at 120 Hz

38. Testbed #1 Quantify application performance vs. complexity tradeoffs
Extend our real-time DSL testbed (deployed in field)
Integrate ideas from multiple narrowband PLC standards
Provide suite of user-configurable algorithms and system settings
Display statistics of communication performance
Investigate
Adaptive signal processing algorithms
Improved communication performance 2-3x

39. Message-Passing OFDM Receiver
RT controller
LabVIEW RT
data symbol generation
FlexRIO FPGA Module 1 (G3TX)
LabVIEW DSP Design Module
data and reference symbol interleave
reference symbol LUT
43.2 kSps
8.6 kSps
zero padding (null tones)
generate complex conjugate pair
103.6 kSps
256 IFFT w/ 22 CP insertion
368.3 kSps
NI 5781
16-bit DAC
10 MSps
BER/SNR calculation w/ and w/o AMP
FlexRIO FPGA Module 2 (G3RX)
14-bit ADC
sample rate conversion
400 kSps
time and frequency offset correction
256 FFT w/ 22 CP removal
FlexRIO FPGA Module 3 (AMPEQ)
null tone and active tone separation
184.2 kSps
51.8 kSps
ZF channel estimation/ equalization
AMP noise estimate
Subtract noise estimate from active tones
data and reference symbol de- interleave
Host Computer
LabVIEW
43.1 kSps
256 FFT, tone select
testbench control/data visualization
differential MCX pair
Example Input Noise
Resource Utilization