Statistical Time-Domain Analysis of Equipment in Low-Voltage Networks

—The increased use of non-linear equipment results in conducted electromagnetic interference problems. As an example, static energy meters, used to measure the energy consumption in households, show misreadings due to pulsed currents from a speed controlled water pump or dimmed lighting equipment. To determine the existence of similar equipment in low-voltage networks, this letter surveys current waveforms drawn by equipment in the low-voltage network. These waveforms are analyzed and compared to critical waveforms from previous research that caused static energy meter interference. In this way the extent of critical situations in a real-world environment is determined. The equipment survey is performed using a multi-channel time-domain measuring method able to capture the voltage and currents synchronously. Then using a time-domain parametric model a statistical overview of parameters such as the charge, crest factor, peak value, pulse duration, and rising slope is given. The survey shows the existence of a large variety of non-linear signals, and 43% contained critical parameters related to static energy meter interference.


I. INTRODUCTION
I N LOW-VOLTAGE networks there is an increased use of equipment with a non-linear behavior compared to traditional linear equipment, which resulted in several conducted electromagnetic interference (EMI) cases.For example, misreadings of static energy meters, used to measure the energy consumption in household situations for billing purposes, are found due to dimmed lighting equipment of light emitting diode (LED) and compact fluorescent lighting (CFL) technology [1], [2], and a speed controlled water pump [3], [4].Accordingly, maximum experimental errors of 2675% are found, resulting from fast changes in pulsed currents that have a small pulse duration [5].The presence of similar equipment and resulting static energy meter interference is shown in [6].Other EMI problems arise in power line communication (PLC) [7], where the current pulse duration affects the communication [8].A study of Bas ten Have, Tom Hartman, and Niek Moonen are with Department of EEMCS, University of Twente, 7522 NB Enschede, The Netherlands (e-mail: bas.tenhave@utwente.nl;tom.hartman@utwente.nl;niek.moonen@utwente.nl).
Frank Leferink is with the Department of EEMCS, University of Twente, 7522 NB Enschede, The Netherlands, and also with THALES Nederland B.V., 7554 RR Hengelo, The Netherlands (e-mail: leferink@ieee.org).the harmonics generated by household appliances is provided in [9], which focuses on the harmonics until 2.5 kHz (50th harmonic), instead of the frequency range till 150 kHz that is of interest for conducted EMI problems, as there is a lack of civil standards in this frequency range.In this letter, the extent of EMI to which static energy meters are exposed to is investigated by the analysis of an extensive survey of current waveforms drawn by typical household equipment.In this way, the existence of equipment with critical time-domain parameters related to the interference of static energy meters is researched.This is done using a time-domain EMI approach [10], capable of synchronous multi-channel measurements [11].These waveforms are analyzed in the time-domain using a parametric model as described in [12].And a statistical overview is presented dividing equipment per category according to the IEC 61000-3-2 [13].In the end the results are compared to critical waveforms that resulted in static energy meter interference.It is found that the surveyed equipment is non-linear and could exceed the test signals that the immunity standard IEC 61000-4-19 [14] envisions.Meaning that a realistic system is Take-Home Messages:

Visual Summary:
• The majority of the surveyed equipment in lowvoltage networks behaves non-linear.• For 43% of the waveforms, parameters inside the critical range related to EMI cases are found.• Combining these non-linearities in a realistic system could exceed the limits immunity standards as the IEC 61000-4-19 envision.
combining non-linearities and thus cannot be described by a frequency domain test, i.e. the response of the system, as [15] already pointed out.

II. ANALYSIS OF CRITICAL PARAMETERS RESULTING IN METERING INTERFERENCE
Current waveforms associated with static energy meter interference are non-linear and have a high pulsed nature [5].This is visible in Fig. 1, where current waveforms from a speed controlled water pump that interfered with static energy meters are shown.In general these waveforms have a very short pulse duration, high rising slope and high crest value.The time-domain parameters of such waveforms are correlated to metering errors in [12], and despite the fact that no univocal correlation was found, a combination of parameters did correlate to metering errors.The critical ranges in which the interference occurs are shown in Table I.In this regard it is of interest to find the existence of equipment with waveforms that have similar parameters.The variety of equipment surveyed is subdivided into different categories according to the IEC 61000-3-2 [13].This standard specifies the limitations of harmonic currents injected into the public low-voltage network for equipment with a rated current up to 16 A.The four categories have different emission limits, and are listed below: • Class A: normal usage: -Household appliances.
• Class B: very short usage: -Portable tools.
• Class D: normal usage, but special current waveform: -E.g.personal computers, monitors and receivers.
For equipment with a rated power of 75 W or less, other than lighting equipment, no emission limits apply.This equipment is of particular interest, because the waveforms that result in metering errors also have a low power consumption.This equipment is treated as Class A, B, C, or D depending on its usage.In addition, when combining a plug-in dimmer with (a series) of lighting equipment no emission limits are applicable, and a combination could generate highly distorted currents, that might result in metering errors [1].Such combinations are treated as Class C equipment when found in a real-world situation.Further, the performance of equipment that is used (or installed) for a longer period may reduce because of ageing of electrical components [16], this might effect the current emissions.Therefore, we focus on installed equipment rather than brand new equipment, which corresponds to its use in an everyday household situation.

IV. METHOD
To cover equipment from the different categories, measurements are performed at different residential locations: an office environment, a hardware store, and at a boiler manufacturer, all throughout The Netherlands.During the measurements the mains supply of the test location is used.These might differ slightly from each other, which portrays the realistic on-site environment where the static energy meter is located and is therefore not a problem.After the mains plug a breakout box is connected, such that the currents and voltage can be measured safely according to Fig. 2, followed by a connection point for the measured equipment, Fig. 3.The voltage and currents are measured synchronous using a multi-channel time-domain EMI approach [11].Pico Technology current probes TA189 within 0.5 dB in the frequency range up to 200 kHz, calibrated according to [17], and a Pico Technology TA043 differential voltage probe with a frequency range up to 100 MHz, are used and these are connected to a Picoscope pc-based oscilloscope.At least ten cycles at mains frequency are captured, which is equivalent to 200 ms, in accordance with IEC 61000-4-30 [18], and a sampling frequency of at least 1 MS/s is used.The waveforms are post-processed to analyze the time-domain parameters: charge, crest factor, peak value, pulse duration and rising slope, using the parametric model as described in [12].The parameters are presented statistically using the Turkey's boxplot method [19], which visually summarizes the data using the following components: the median (Q2), the lower (Q1) and upper (Q3) quartile, the minimum (min), and maximum (max).Points that have a value of 1.5 times the interquartile range from the median are marked as outliers and are visible with a dot in the graphs presented in Section V.

V. RESULTS AND DISCUSSION
During the survey a total of 484 different equipment situations are measured, were Table II provides an overview of the surveyed equipment per category, and the number of appliances with critical parameters according to Table I.A statistical overview of the charge, crest factor, and peak value is given in Table III, the parameters inside the critical ranges are highlighted in red.A detailed visualization of the pulse  duration and rising slope is given in Fig. 4, as those parameters are found to be more critical in [12], the critical range is indicated in red.And 43% of the equipment contains critical parameters.For the slope especially, several outliers, with a high slope in the critical range, are found in all categories.For the equipment surveyed in Class A (normal usage household appliances) 29% was linear equipment, were in some of these cases the current was phase shifted with respect to the voltage, which will contribute to the reactive power in the circuit.All this linear equipment did not include critical parameters.For the non-linear equipment in Class A, in a couple of cases the slope is higher than 0.1 A/µs, which resulted from a phone charger, washing machine and speed controlled water pump.For the phone charger the other parameters were far outside the critical ranges, as the pulse had a low peak value and charge.The pulse of the washing machine was quite significant, Fig. 5, and the crest factor and slope are inside the critical range related to known static energy meter interference.The extremity of the time-domain emission of the water pump and the resulting existence of static energy meter interference was already shown in [3].
The waveforms generated by portable tools in Class B contain a lot of charge compared to the other categories, and as a result consume more power.Consequently, the median peak value is higher and the pulses are usually wider in comparison with the other categories.Due to this higher pulse duration, the crest factor of these waveforms is lower.Also no slopes in the critical region were found.The lighting equipment in Class C has slopes inside the critical range which mostly occur in combinations with dimmers.Cases with narrow pulses with a peak value above 1 A and a comparable wave-shape to the waveforms resulting in static energy meter interference in [1] were found.Furthermore, a case occurred in which transient pulses occur in dimmed lighting equipment, these have a very short pulse width and high slope, Fig. 6.The peak value and therefore also the charge however are low, making their influence on the meter readings questionable.However, [4] has also shown the existence of transients in dimmed lighting equipment resulting in static energy meter interference, but those transients had a 50 times higher peak value.
For the equipment in Class D the crest factor indicates the existence of highly non-linear waveforms.However, this equipment has a lower power consumption, indicated by a low charge value, and also the max peak values are low compared to the other categories.Furthermore, no extreme slopes in the critical range are found.
The presented survey evidences the existence of equipment with a non-linear behavior including parameters that could be critical in relation to static energy meter interference.Therefore, a realistic system that is combining these non-  linearities could exceed the test signals immunity standards as the IEC 61000-4-19 [14] envisions.Such tests are only valid if the system can be considered as linear time-invariant, however as the results show, this is clearly not the case, this issue was already pointed out in [15].

VI. CONCLUSION
The surveyed equipment has shown that 43% of the equipment has parameters inside the critical range related to possible static energy meter interference, and in the majority of cases the equipment is non-linear.Thus a realistic system that uses a static energy meter is combining these non-linearities and could exceed the test limits immunity standards as the IEC 61000-4-19 envisions.The survey showed waveforms with fast inclining slopes above 0.1 A/µs, which resulted from a phone charger, lighting equipment with dimmers, a washing machine and a speed controlled water pump.Both the washing machine and speed controlled water pump showed several parameters inside the critical range related to static energy meter interference.The power consumption of the charger however was low, resulting in low peak values compared to the others.For the lighting equipment both waveforms with fast rising narrow pulses and high peak values, as well as waveforms including small transients with low charge and peak values were observed.
Manuscript received June 21, 2021; revised August 19, 2021; accepted September 19, 2021.This project 17NRM02 MeterEMI has received funding from the EMPIR programme co-financed by the Participating States and from the European Union's Horizon 2020 research and innovation programme.(Corresponding author: Bas ten Have.)

Fig. 1 .
Fig.1.Non-linear currents causing static energy meter interference, generated by a speed controlled water pump at different speed control levels[3].

Fig. 4 .Fig. 5 .
Fig. 4. Statistical overview of the pulse duration and rising slope of the surveyed equipment.

TABLE I
[12]itical ranges of time-domain parameters resulting in static energy meter interference, as described in[12].

TABLE II .
[13]view of different surveyed equipment and equipment containing critical parameters, per category according to IEC 61000-3-2[13].

TABLE III .
Statistical overview of the charge, crest factor, and peak value of the surveyed equipment.