This section presents the theoretical framework on which the presented work is based as well as the materials used in it.
2.1. Theoretical Frame
The previous documents mentioned regarding single-cable transmission in
Section 1 considered frequencies of transmission above 50 MHz (higher than the ones we worked with in this first demonstration stage for our PV–PLC system), whose current distribution is expected to be flowing superficially or in a thin layer close to the surface of the cable (skin effect), with propagation described to be carried out by a surface wave whose phase velocity is c (the speed of light). Our direct measurements of the phase velocity over a line of single PV cable at a frequency of 20 MHz showed speeds around 0.6 c, closer to what is observed in typical coaxial cables [
14], even when the propagation mode is different in our single-cable medium (TM) than the one in coaxial cables (TEM). The lower frequency used in our single-cable support suggests that the current distribution flows in a thicker layer inside the conductor, and the influence of the material lowers the phase velocity with respect to the previously mentioned Sommerfeld surface wave. As a result, the single-cable analogy applied to the PV plant cable cannot allow us to assume the same propagation mode described in
Section 1, and since our aim with this document is not to establish a theoretical background but to develop a practical approach for implementing a working PLC system specific for PV plants, we assumed the measured phase velocity of 0.6 c and tried to start from a simple model to characterize our system.
The influence of the topology of the physical wiring in the previous literature on PLC systems for PV strings is not usually discussed, probably because the low-frequency carriers proposed do not show limitations related to higher frequencies, such as interferences or resonance issues. Some previous works use an open line topology [
7,
15,
16] and some others propose a current loop topology [
17]. The frequencies are normally in the order of 100 kHz or less, that is to say, considering the assumed phase velocity (0.6 c) wavelengths are around 1800 m, much bigger than the usual length of the string. With this work, we try to explore the possibility of setting the carriers of higher frequencies (from 1 MHz on), analysing the effects of the increased frequency and offering a practical solution, which inevitably leads to considering the physical topology of the wiring.
In a single PV string, the power wires starting from a combiner box return to the same box after connecting the modules in series. In essence, it is easy to establish a circular communications path by means of a capacitor connected between the two cables arriving at the combiner box and to bypass capacitors connected through each PV module. This allows for a cable loop configuration to be a communications physical support for a single string. The choice of a closed-loop configuration for the signal path has several advantages for communications purposes. This topology sets all the points along the circular line at the same level regarding attenuation, so differences in reception levels are reduced to a minimum and it allows for a resonance condition to be established for the signal along the loop, since there is a natural reactive impedance associated with a cable loop depending on the relationship between the wavelength and the physical length of the loop, which can be compensated for by some added lumped reactive impedance, leaving only a low-resistance path for the signal.
Adding the ring topology proposed to the aforementioned assumption about phase velocity = 0.6 c, a simple propagation model can be considered. Our starting point is the schematic in
Figure 1.
The cable ring of perimeter d represents a simplified sketch of the signal path, including the PV cable line and bypass capacitors (with enough capacitance for presenting a very low impedance at our working frequency), where the I/O impedance of the transceivers located along the ring is neglected for a first approximation. It is excited by a current generator (which in practice could be the secondary binding of a transformer). A general expression of the current wave along this loop is:
Representing the sum of two waves moving in opposite directions along the ring, where I is the current amplitude with its time dependency:
Here, x is a coordinate indicating the distance along the loop measured clockwise from the generator position and k is the wavevector depending on the frequency (f) and phase velocity (0.6 c) and A and B are constants (which could be complex) to be determined.
The boundary conditions imposed are related to the continuity of the current at the generator thus:
leading to
In general, the solutions of this system of equations lead to (complex) values of A and B different from zero and thus to a stationary wave pattern for the current distribution. These solutions are not desirable to implement a communications system because stationary waves give rise to maximum and minimum current amplitude points along the loop, which means different reception levels since a transceiver could be located at any point in the ring. For the case where
where n is an integer value, this system of equations is incompatible and there is no solution, representing a destructive interference between clockwise and anticlockwise waves. The optimal solutions for our purpose are those where A = 0 or B = 0, which leads to a travelling wave solution of the form:
not presenting maximum or minimum current amplitude points. To obtain these kinds of solutions, it is mandatory that:
that is to say, the perimeter of the loop is an integer multiple of the wavelength.
However, even when the above condition is satisfied, there are stationary wave solutions (A and B ≠ zero) satisfying only the extra condition A + B = 1. In order to determine the actual spatial current distribution over a typical PV cable loop, we built a setup comprising 20 m of cable from our real test PV plant (Figure 6, show later) and measured and assessed the string as one single conductor line travelling in a straight fashion from the positive lead of one signal generator to the end of the PV array arrangement where it then turned around, coming back as a loop at a distance of one metre apart from the initiated point at the positive lead, finishing in the negative lead of the signal generator. The topology of the circuit is thus a loop-like one, but the shape is elongated from a circular one. The generator was adjusted for a sinusoidal signal of 9.142 MHz (determined as the signal corresponding to a wavelength equal to the loop length, fulfilling the above condition) and 10 volts of amplitude. The validation of the system in short wire installations is interesting, since for longer installation lengths resonance can be achieved at the same frequency by means of higher-order modes, where more than one wavelength is present in the line. Essentially, if it works for our 20 m loop, it will work also in loops which are 40 m, 60 m, etc., and the fine adjusting for intermediate lengths can be easily achieved by the tuning circuit designed (see
Section 2.2).
The current measurements were made in a low-invasive fashion at different points along the loop, sensing the AC magnetic field associated by means of a toroidal ferrite core surrounding the cable, with a 30-turn enamelled cable wiring. This way, an induced AC voltage proportional to the AC current in the line can be measured with an oscilloscope across the terminals of the wiring. The high input impedance of the oscilloscope guarantees that a very low current will flow through the wiring, so a negligible inductance was added to the loop during measurements.
Figure 2 shows the results of these measurements taken every metre along the loop, and it shows a clear stationary wave pattern. These results agree with the literature about the current distribution on loop antennas, the theory of which is close to the one here considered.
The difference between the maximum (100%) and minimum (21%) current amplitude levels from the measurements represents 13.55 dB, which can be managed easily by an automatic gain control (AGC) within receptors. There are two possibilities for fully exploiting the advantages of a resonant loop for communications avoiding significant differences in the signal levels along the loop: (I) Find a simple way to excite only the travelling wave mode over a one-wavelength ring (optimal). (II) Work at carrier frequencies low enough in a fashion that the loop length is a fraction of ¼ of the wavelength or less. This is the slower option, but it still leads to carrier frequencies over 2 MHz (assuming a propagation speed of 0.6 c and a loop length of 20 m), significantly higher than the average previously reported.
Regarding solution I, some authors have faced the theory of travelling wave solutions for loop antennas [
18], compared with standing wave solutions [
19], and even have determined the necessary conditions for exciting the travelling wave modes [
11]. All this work, however, is focused on antenna design and consequently looks for the best radiation properties. For our application, radiation losses are not desired, since they mean power losses and could affect the electromagnetic compatibility of the system. Fortunately, the drastic reduction in the loop area due to the particular shape of our setup will prevent the system from great radiation losses, since radiation integrals obtain contributions of opposite current elements much closer to each other [
20]. The travelling wave or quasitravelling wave mode of propagation for our system could be achieved by fulfilling some of the conditions exposed in [
18,
21].
However, these conditions are not easily applicable in a practical environment, so the insertion in series of a nonreciprocal device is proposed. These kinds of devices will attenuate the waves travelling in one direction, leaving unaffected ones travelling in the opposite direction, leading to a travelling wave propagation mode. Some of the possible practical devices that could accomplish this task are magnetic-circulator-based isolators or unity gain amplifiers referenced to Earth in such a way that waves arriving towards the output will find very low impedances to Earth, being attenuated, but the ones arriving towards the input will find very high impedances and will progress to the output almost unaffected. In this work, we focus on resonance control, and isolator insertion is left for future research.
Regarding solution II, if we work at wavelengths four times the length of the loop or bigger, the phase changes along the loop in
/4 radians maximum. For a loop length of ¼ wavelength, the general solution for its current wave is:
where kx goes from 0 to
/2 along the loop length, so the maximum change in the current wave amplitude is:
This value is small enough to avoid AGC systems in receptors and simplifies the electronics, allowing carrier frequencies over 2 MHz as explained before. This carrier (depending on the modulation system) can lead to baud rates around 200 kbps, which are higher than the previous baud rates reported in the literature for PLC within PV plants, so even when this mode of propagation represents the lower baud rate for our model, it still supposedly should provide an improvement on the speeds previously reported.
In addition, the resonance condition can still be achieved for this case because even when the impedance variations related to auto interference are almost not present, there is an inductance associated with the line and reactive impedances representing the transceivers (capacitive or inductive depending on the coupling chosen) that must be compensated for with a lumped element in order to push the loop to resonance. For the higher-frequency (lower-wavelength) option, as a starting point, we can consider a model of signal propagation along a closed loop similar to the ones previously mentioned regarding loop antennas, which show series resonances (impedance close to zero) at frequencies whose wavelengths are integer fractions of the loop lengths [
20]. In this way, a constructive interference gives rise to a spatial resonance, and a maximum in the signal amplitude is observed. Since one key goal of a communications system is the integrity of the signal, we must satisfy the above condition in our cable loop in order to work at an optimal point with the best SNR possible.
For a chosen carrier frequency, a loop length is fixed to fulfil the above condition; however, the length of the loop is an imposed parameter depending on the physical dimensions of the installation, and therefore it is necessary to find a way to adjust the natural loop resonance to match the frequency of the carrier. The insertion of a coil in series with the cable loop has the effect of increasing the electrical length seen by the signal (length expressed as a wavelength multiple), that is to say, the loop will resonate at lower frequencies. The opposite of this behaviour is seen with the insertion of a capacitor. As such, the insertion of a reactive component in series with the loop could be used to perform the matching between the carrier and loop resonance frequencies. This effect is shown in
Figure 3 from our measurements over a typical PV string cable loop 20 m long with a vector network analyser (VNA) connected to the loop, where the resonance condition is recognized by the minimum in the modulus of the S11 scattering parameter (maximum power sent to the loop) and a sudden change in its phase.
Figure 3 shows curves around the first resonance frequency (9.192 MHz without compensation) corresponding to the connection in series of a coil, three values of capacitors and the raw cable.
These measurements have been used to determine more precisely the phase velocity in our PV cable. Since the first cable resonance (9.192 MHz) corresponds to one wavelength in the 20 m long cable:
2.2. Materials
Even though a preinstallation length compensation is feasible (installing a fixed reactive series component), employing automatic adjustable tuning would more so be convenient in confining the transmission around an optimal point, in order to account for the inevitable occurrence of small variations in the loop (for example, the addition of one more PV module to a string, which would increase the loop physical length). For this purpose, a simple tuning circuit was designed which was able to show either capacitive or inductive impedance by means of a control voltage. The basic variable component used was a varicap diode whose capacity could be adjusted depending on the inverse voltage applied. The circuit is shown in
Figure 4a, and it is composed of a toroidal transformer with the primary (L2) connected to the cable loop and the secondary (L1) connected to the variable capacitor (C) (voltage controlled). The actual implementation circuit for the voltage-controlled capacitor is shown in
Figure 4b.
The network analysis of the circuit in
Figure 4a leads to an impedance seen from the primary port:
where the mutual inductance coefficient in the transformer is expressed by:
and
is the coupling coefficient, showing a parallel resonance at
and a series resonance at
This impedance has two inductive regions, (
and
, and one capacitive region:
. In the centre of the capacitive region with a good coupling in the transformer
, we have
and
, so
and our circuit works as a capacitor multiplier by the factor
, which can be very useful for loops requiring high-capacitive compensation with values higher than the maximum achieved by the varicap diode, for example, for loops with high inductance in series (the inductive coupling of the transceivers).
For loops with low series inductance (the capacitive coupling of transceivers), the usual length of a typical PV string and carrier frequencies over 5 MHz, our circuit is better used in the region around the higher frequency of resonance
, where the impedance is close to zero, and for frequencies lower than
the impedance is capacitive, and for frequencies higher than
the impedance is inductive. Around this region and
, the impedance can be expressed as:
equivalent to a series LC circuit with
and
resonant at
. Since the capacitor
can be dynamically adjusted, connecting this circuit in series with the cable loop, we can add series inductance, capacitance or none of them depending on what is required by the loop, and the dynamic adjusting range can be selected with the value of
since it is a common factor in the expression of Z around
. The possibility of the capacitance control varying a biasing inverse voltage on the varicap diode allows us to implement a microcontroller-based tuning system that will also require some kind of detection of the resonance condition in the loop.
The carrier amplitude through the primary coil in the circuit in
Figure 4a will show a maximum if the loop is pushed to resonance, so if this signal is used as an input for a voltage doubler detection circuit (
Figure 4c), we will obtain at the output a DC voltage proportional to the amplitude of the carrier in the loop and the resonance can be detected as a maximum in the DC voltage output. Finally, automatic resonance adjusting can be implemented by sampling the detector DC output with an analogue-to-digital converter (ADC) integrated in the microcontroller (MCU), where the convenient firmware will search for a maximum outputting a DC voltage sweep towards the tuning circuit.
Figure 5 shows the real implementation of the circuits proposed, with all the components labelled.
Figure 5a shows the front side of the board and
Figure 5b shows the back side.
The PV plant used is shown in
Figure 6. This plant is located on the facilities of the Duques de Soria campus of the University of Valladolid. The campus is in the city of Soria, Spain. The modules used in this article for the communication line are those on the upper row. Each module has the characteristics shown in
Table 1.