Inaccurate energy meters

[Janma]

This morning I came across this interesting article : https://www.sciencedaily.com/releases/2017/03/170303180139.htm

I effectively have to energy meters at home. The old disk type meter (municipal meter) and a grid tied limiter and I've noticed that their measurements don't always match 100% At first I thought it was a calibration issue. From the study it seems a much bigger problem. I wonder how pervasive this problem is with the energy meters in SA, especially with the increase in popularity of LED light bulbs.

 

[___]

Okay, I'm going to let you in on a secret. Every electronic meter on the market uses an estimate. What it has to do is measure both the voltage and the current, then multiply the two, and that gets you a power figure. Sounds simple enough, but you also have to consider power factor. Even that is easily accounted for as long as you can make certain assumptions about what the waveforms look like, and this is where things are going squivy at the moment: The current waveform is often severely distorted.

So what the meter has to do now is measure the voltage and current (which it will do by taking a number of samples per second), then it has to multiply those values to get an instantaneous power value, and then it has to integrate (as in sum the area under the graph, if you actually plotted it) of the resulting values. If you had calculus at some point in your career, you may remember the dt term, which is essentially a time-step that strives towards zero (and then you take the limit of all that, it's been a decade since I've done this). Integration is all fine when you're dealing with well-behaved smooth waveforms, but when you're not, you have to use some kind of numerical approach, usually by fitting trapeziods over small time steps.

The better your integration algorithm, the closer it will be to the actual value. It will never be perfect, but you can get close.

I think what is now being discovered is that a lot of meters out there employ hopelessly naive algorithms to do this.

Edit: Or perhaps they are based on microcontroller tech that simply can't sample that fast or don't have the CPU to do the calculations fast enough, because when you're rolling out a couple hundred-thousand meters, you want the cheapest one, right?

[Janma]

@Plonkster : The sampling frequency may be low and that could cause a error. That does not explain why Hall effect current sensors under estimated the energy and why Rogowski coils greatly over estimated the energy. (Both would make use of similar integrator circuitry)

I suspect problem is in the frequency response of the sensors  and the approximations they use to calculate the energy.  Rogowski coils gives a voltage output that is proportional  to the strength of the current and its frequency (Integrate over dI/dt). I think the designs they use assume that the frequency will only be around 50Hz, but that is not always the case. LEDs, Energy saving Light bulbs, electric motors, dimmers and switch mode power supplies all inject much higher frequencies (Harmonic distortion) into the grid. These harmonic distortions, because they are at higher frequency cause the approximation they use to fail.

The approximation they most likely use is  V_sensor = K * I_rms * F   or I_rms = V_sensor/(K*F)

Where K = a constant specific for the sensor

F = 50 Hz (assumed)

This approximation works well if the frequency is 50 Hz, but it over estimates the current contributed by the second harmonic by a factor of 2 and that for the 10th harmonic by a factor of 10!

In the old days this was not such a big problem because there were no LED, Energy saving light bulbs or switch mode power supplies etc. The big electrical energy consumers in a household were Geysers, Heating elements, stoves, incandescent light bulbs etc. none of which inject any significant harmonic distortion.  Things have changed though. Take my home for example, I've replaced all the incandescent light bulbs with LEDs. I use a solar geyser and gas stove. Other than the kettle and toaster almost all appliances in my home either have an electric motor or switch mode power supply, all of them inject harmonic distortion to some extent.

I'd be interested to measure the harmonic distortion at my home, just to see what the impact of this would be if I were to move from the mechanical rotating disk meter to a new electronic prepaid meter.

 

[___]

56 minutes ago, Janma said:

That does not explain why Hall effect current sensors under estimated the energy and why Rogowski coils greatly over estimated the energy

Well, I had to google Rogowski cause I had never heard of it until that article. Only then did I learn that it is a kind of current transformer.

The hall-sensor devices probably take samples and apply a tiny bit of smoothing (or averaging), so the rough edges caused by bad appliances/lights are ignored so to speak, so they will measure low. That makes sense. I think you're right that the real challenge is figuring out why a CT-based meter would measure high.

As part of my job, I deal a lot with algorithms involving sensors that read the human body, and smoothing and some kind of bandpass filter is almost always the first step to making the data more unwieldy... :-)

[Chris Hobson]

https://www.utwente.nl/en/news/!/2017/3/313543/electronic-energy-meters-false-readings-almost-six-times-higher-than-actual-energy-consumption