Improvements to power line communications for AMR
Commercial two way power line communication (PLC) systems for automatic meter reading (AMR) are based on relatively crude signalling techniques and sometimes unsuitable frequency ranges. This situation is a serious impediment to the rollout of optimally cost-effective AMR.
The continually reducing cost of processing power has lowered the barriers to the use of sophisticated methods. We have witnessed this in the cell (mobile) phone world, where updating signalling schemes has vastly improved the performance (users per unit bandwidth) and reduced per user system cost.
Perhaps progress has been delayed in utility-owned PLC because worldwide deregulation of the industry has produced uncertainty and therefore reluctance to invest in new technology. However the need for low cost two-way communication is becoming more important, because it is required for effective demand management – now vital in some overloaded areas to avoid the huge costs associated with interruption of supply.
Let us look at some of the technical aspects behind improvements to PLC for AMR.
WHAT’S REQUIRED FOR AMR COMMUNICATIONS?
In the past ten years there has been a lot of ‘hype’ associated with broadband power line communication for Internet service delivery. It may or may not be a rewarding area for utilities, but broadband is unnecessary for the very limited amounts of data associated with AMR and related functionality. Broadband over power lines (BPL) is often advertised as having big spin-off advantages for utilities because AMR and other utility-based functions can be hung off it.
This is true, but the cost of rollout of a system capable of broadband performance is huge. The associated high frequency signals (tens of MHz) simply do not travel far down the power network. This necessitates addition of repeaters and couplers, with associated installation and maintenance costs. BPL is a red herring as far as PLC for a utility’s use is concerned.
What is required is a system which uses the power network as it stands, with no modification whatsoever. Further, a really useful system must have rock bottom meter end cost. These two factors dictate the signalling frequencies and, less directly, the modulation, channel access and coding schemes which are best used.
SIGNALLING FREQUENCY RANGE
In Australia and Europe some success has been achieved using signalling frequencies around 100Khz. These frequencies do not pass over distribution transformers, and so the systems are limited to transmission of data over the low voltage (LV) part of the network. Signals are collected at data concentrator nodes and sent over a public data network such as GPRS. Economic viability is possible with typically 60 to 100 dwellings per distribution transformer, the cost of the data concentrator being amortised over a sufficient number of meters.
With the different network topology in America it is vital that the signal is not stopped by the transformer. Load control systems in Australia use ‘ripple signals’ between 1050Hz and 167Hz, and our experience at La Trobe confirms that only frequencies below about 2250Hz will adequately pass over distribution transformers and still be detectable above the noise. (See the frequency response plot in Figure 1).
Use of these low frequencies enables a whole population of perhaps five thousand meters to communicate right back over the LV and MV lines to their local zone substation from where a low cost backbone communications link is generally available.
Perhaps a good way to consider this aspect is to look at what has been used in the past, with the associated advantages and disadvantages.
Sinusoidal voltage signalling
The ripple control signals previously mentioned are keyed sinusoidal audio frequency signals added to the mains supply voltage. The utility adds these signals at the zone substation for the purpose of broadcast downstream communication. The typical signal level used is 10 Volts, so the substation transmitter is a large device.
Let us consider for a moment using the same scheme – that is modulating the data onto sinusoidal voltage signals – for upstream transmissions. The signal level produced by the meter end transmitting device is necessarily small to start with, and miniscule by the time it reaches the substation. If the received signal is miniscule, the receiver bandwidth must be ultra-narrow to successfully detect it in the presence of noise. It is an unbreakable rule of nature that the minimum bandwidth a fixed level signal occupies is proportional to the modulating data rate, so an ultranarrow band system is equivalent to an ultra-low data rate system.
It is true that the data rate requirement for AMR and associated functionality is small. However, in the state of Victoria in Australia, the government has recently mandated half hour interval metering for residential customers, so that price signalling at peak load times can be implemented. This means 48 meter readings per meter per day are now required. Although there is scope for data compression, the amount of data is far above the one meter reading every three months we have been used to.
These new requirements put a lower limit on the data rate any practical system must produce and an ultra-narrow band system is unlikely to be able to cope.
Current pulse signalling
To transmit information over a channel we modulate an electrical parameter which can be detected at the other end. It is as valid to use current for signalling as voltage, and producing a large current signal at the meter is relatively easy. This current signal is not attenuated by a long line with loads attached to the same extent that a voltage signal is.
One commercial system uses a triac to switch the current in a load. With an inductor as the load, dissipation is minimised yet the current drawn (our signal) can be made very large indeed. 60 to 70 Amp current pulses were being used for the system described.
The big disadvantage of this simple transmit circuit is that it can only produce pulses centred on the mains voltage zero crossings. This results in disappointing performance for two reasons. First the zero crossings are limited at 50 or 60 per second per phase, so there is no scope for a large number of transmitters to use the channel simultaneously. Secondly the frequency content of the signalling waveform is similar to the frequency content of the interfering signal (the mains load waveform). Both contain 50/60Hz and also mostly odd harmonics of 50/60Hz. The result is that a previously recorded mains cycle must act as a reference to determine the presence or absence of a pulse. This means the system relies on the mains load being at least quasi-static to avoid errors in reception.
Adding a capacitor in series with the switched inductor (see Figure 2) improves flexibility. When the triac fires, the circuit gives a half sine pulse of current with a width determined by its resonant frequency. When the triac is fired again (if the mains is the same phase) we get a current pulse of the opposite polarity. With appropriate care in the timing of the firing of the triac it is possible to output current pulse trains with almost any fundamental frequency below that of the circuit’s resonant frequency.
Now the restrictions of the previous current pulse circuit are lifted. Inter-harmonic frequencies can be generated which are much easier for the receiver to distinguish from the mainly harmonic noise. In addition it becomes possible for a great number of users to share the channel, each modulating the phase of a different transmit frequency. Such a system has been proved to work very satisfactorily in Australia but it was too expensive for the market at the time, necessitating bulky inductors and an additional high voltage capacitor.
Use of this type of current pulse signalling with a more efficient channel access scheme (based on spread spectrum techniques) promises improved performance with reduced signal level. This is what we are working on at La Trobe University.
At La Trobe we have demonstrated, using our on-site power station and its MV/LV distribution network as a test platform, that a typical PLC channel has a surprisingly high maximum theoretical capacity of at least some tens of times the capacity utilised by existing ‘over the transformer’ PLC systems. This means that data rates suitable for interval metering AMR are theoretically possible using low frequencies that pass over transformers.
Making best use of limited bandwidth and transmit power is not a problem confined to PLC. Many people have come up with wonderfully efficient solutions; perhaps the most extreme example is communication with deep space probes. Even though the application is ‘world’s away’ from PLC, some developments are directly applicable. Efficient channel coding schemes are one example and are beginning to be looked at by the PLC industry.
We hope the gap between what is theoretically possible and the current situation will be closed using recently developed channel access and coding schemes. If successful, the new Australian infrastructure being put in place now to serve us for the next 20 years need not be hobbled by using techniques already 20 years out of date.