Through the Soil Long Range Wireless Power Transfer for Agricultural IoT Networks

Increasing the spatial and temporal density of data using networked sensors, known as the Internet of Things (IoT), can lead to enhanced productivity and cost savings in a host of industries. Where applications involve large outdoor expanses, such as farming, oil and gas, or defense, large regions of unelectrified land could yield significant benefits if instrumented with a high density of IoT systems. The major limitation of expanding IoT networks in such applications stems from the challenge of delivering power to each sensing device. Batteries, generators, and renewable sources have predominately been used to address the challenge, but these solutions require constant maintenance or are sensitive to environmental factors. This work presents a novel approach where conduction currents through soil are utilized for the wireless powering of sensor networks, initial investigation is within an 0.8-ha (2-acre) area. The technique is not line-of-sight, powers all devices simultaneously through near-field mechanics, and has the ability to be minimally invasive to the working environment. A theory of operation is presented and the technique is experimentally demonstrated in an agricultural setting. Scaling and transfer parameters are discussed.

many industries. Sensor installation throughout the industrial process plays a fundamental role in these tools. The number of sensors that can be installed is limited by the following two primary factors: 1) the cost (including installation/maintenance) of the sensor; 2) the power source used by the sensor. Recent advances in electronics/manufacturing have effectively solved many of the sensor cost challenges. However, there are few low-maintenance/low-profile solutions to delivering power to a multitude of sensor systems. Difficulties in power delivery are only amplified when applications require monitoring in remote and expansive outdoor environments. This is specifically challenging in industries such as farming, where large tracts of land are usually far from any electrical supply. Weather-dependent sources such as photovoltaics and wind turbines, or the continual replacement of batteries are the only real solution to this problem, all of which do not support Internet of Things (IoT) scaling. These conventional power methods can be intrusive to the working environment as devices can become entangled in or damaged by equipment/livestock/wildlife. Motorized equipment (such as tractors) must maneuver around solar and wind charging solutions, and as crops, trees, or vegetation mature, they may become completely overshadowed and/or entwined. Thus, contemporary power solutions make the installation of sensing tools time consuming, costly, and in many cases, impractical as the number of sensors are increased.
These electric power supply limitations have led to wireless power transfer (WPT) research that encompasses both near-field and far-field techniques. RF energy harvesting [1] or robotic drones [2], [3] have both been developed for transferring electric power to soil sensors [4], [5], [6], [7], [8], [9], [10]. RF energy harvesting requires the rectification of space-wave signals that do not penetrate far beneath the soil surface and require a substantial time to collect enough energy to perform a function. Drones, either land-or air-based, are generally equipped with a magnetic coupling WPT system that transfers energy to sensors when the drone is positioned directly over the sensor's receiver. A drone solution adds navigation complexities, high-power consumption for locomotion, high operating/maintenance expenses, and limited spatial coverage unless multiple drones are usedcompounding all factors. Neither of these solutions supports wide area IoT scaling, with both suffering from efficiency losses when sensors are buried below 15 cm-as bulk soil is conductive, causing high attenuation of an electromagnetic (EM) wave's magnetic component at the frequencies used.
Long range forms of WPT (LR-WPT) have also been investigated. These rely on the transmission of focused EM space-waves. In this category, the following are the two predominant LR-WPT technologies in use: 1) laser power beaming; 2) RF microwave power transfer (MPT) [11], [12], [13], [14]. Both techniques use far-field EM waves to transfer power. Literature has shown that MPT is more reliable for transmission between fixed installations [11]. Zhu [12] gave a side-by-side summary of present MPT endeavors. These systems either exhibit very low efficiency or have short transmission distances [14], [15]. While some companies are commercializing laser power beaming and MPT-based systems, they are in the very early stages of development with limited prototypes demonstrated [16], [17]. All such systems are line-of-sight dependent and require large, complex rectenna arrays to lower the power density during transmission to improve safety (as the wavelengths used are typically limited to 1 W or lower exposure limits). Neither of these LR-WPT systems would make good candidates for powering expansive IoT networks.
More obscure in present literature is the nonline-of-sight approaches where energy is transferred along the surface of the Earth with a ground wave [18], [19]. Viziv Technologies is a company that claims to be working on ground-wave propagation for power transfer [20], [21]. These systems utilize tall above-ground antenna structures to induce charge oscillations at the surface. To date, no physically realized experiments have been publicly demonstrated by Viziv and based on the data presented in the limited literature, the given distributed electric field (E-field) intensities are much too weak for long-range wireless power transfer [20].
It can be concluded that there currently exists no other techniques beyond line-of-sight far-field radiation that has demonstrated LR-WPT. This manuscript presents a first of its kind WPT concept that utilizes conduction currents "through the soil" (TTS) to transfer power to surrounding devices. The presented system is not line of sight dependent and is observed to be robust in the limited locations tested [22]. The geometry of the TTS system is similar to a water well, offering a possible way to integrate this LR-WPT technique into existing farming infrastructure at very little cost to the user. The theory of operation will be expanded from earlier studies by the authors. A horizontal receiver geometry will be presented and an agricultural case-study using four IoT devices without batteries will be shown as proof of concept. A section on range and efficiency will explore possible methods for future improvements to the system.

II. THEORY AND DERIVATIONS
This section presents methods used to investigate the dynamics of the TTS system. First, a lumped circuit model is developed to identify the parameters that affect efficiency. Next, a concept based on EM field theory is presented for determining the received voltage distribution in the soil as a function of radial distance around the transmitter. This same theory also provides an expression for the soil's impedance, linking the source current to the received voltage at a distance. The theory for the potential distribution is experimentally verified over 0.8 ha in an adjacent hay field. The theory is then combined with the lumped circuit model to arrive at an approximate equation for the system's efficiency with parameters that will serve as metrics for future improvements to the system.

A. Geometry and Circuit Model
The TTS transmitter (Tx) utilized a minimum of two conductors (defined as electrodes) in direct contact with the soil, where one electrode resided at the soil surface and the other resided at a vertical distance below the surface (defined as a vertical geometry). The top electrode was formed from a well casing that was installed around the bore-hole of the Tx. The well casing was made of low carbon steel that was approximately 15-m long. The bottom electrode was a 15-m section of brass tubing (50 mm in diameter), located 75-m below the surface. The brass tubing was connected to a high density poly ethylene (HDPE) tubing of the same diameter that ran from the brass to the surface. A 12-gauge insulated wire was fastened to the inside of the brass tubing at the bottom, and ran to the surface on the inside of the HDPE. Thus, the wire allowed an electrical connection to the bottom electrode from the surface, with the wiring being isolated from the surrounding soil via the HDPE tubing.
The receiver (Rx) for this work was constructed using a horizontal geometry, where both electrodes (0.7-m long) resided at the surface. This is shown in Fig. 1 where an annotated illustration of the geometry is depicted next to a photograph of the experimental system. The horizontal geometry of the Rx was chosen for the ease of deployment and measurement. Future investigations will explore vertical topologies for the Rx, whose placement will become more permanent. A basic schematic of the system and its corresponding circuit is shown in Fig. 2(a) with orange bars indicating the electrodes. Due to soil strata layers being mostly horizontal, currently injected into the ground will follow strata layers of lower resistivity and not spread (or fringe) as far within the surrounding soil medium [23]. This makes horizontal Tx geometries more prone to changes in conductivity due to weather and less effective at transmitting energy over the surrounding area. A vertical Tx structure exhibits greater current fringing since moving charge must traverse all strata layers (regardless of their variations in conductivity). The vertical geometry is therefore a better design choice for the Tx. For the Rx, the only geometric constraint is that its electrodes must reside at different equipotential lines created by the Tx in order for a voltage to be received.
The energy exchange between the Tx and Rx was modeled as a pair of coupled circuits [ Fig. 2(b)]. This model is only slightly different from the current controlled voltage-source model presented in [24]. The difference between the two models is that the coupled circuit allows impedance variations at the Rx to be seen by the Tx, which is expected to occur once the resistive components of Z T X and Z RX are made small. The transfer efficiency (η) of the system can be approximated using this model. Applying loop analysis produces two circuit equations The direction of the coupling is chosen such that current flowing in the received circuit will reduce the reactive component in the supply circuit-similar to a transformer. The received voltage (V R ) takes the form where u e is a lumped "Earth impedance" that is dependent on the distance (r) from the Tx. The potential difference in the soil, (Φ(r)), is derived in the next subsection.
Using (1) and (2), the source current (I S ) and received current (I R ) are solved for yielding: Equation (5) will be used in the next section to validate the model. The input power (P IN ) and load power (P L ) are Dividing (7) by (6) gives the power transfer efficiency (η)

B. Deriving the Potential Distribution
To determine u e , the voltage distribution around the Tx must be derived with respect to the radial distance r. The complex network of resistances, capacitances, and inductances that would be needed to model the soil are difficult to approximate and challenging to measure. Fortunately, the Tx geometry is similar to, and can be approximated as, an electric dipole. By solving for the E-field of this dipole (modeling the medium as a lossy dielectric), it should be possible to integrate the E-field radially to find the potential distribution along the ground. The usefulness of the dipole approximation is that it links source current injected into the Tx to the voltage distribution, mitigating the need of developing a complex impedance model of the soil.
Referring to Fig. 3 for the time-varying electric dipole approximation, the perceived separation distance (d) of the point charges is first modified from a conventional dipole to be the circumference of a toroidal path the current would traverse. (d) becomes where a is the physical separation distance between the electrodes in the soil as shown in Fig. 3. Taking the vector potential ( − → A ) expressed in the frequency domain, the magnetic field intensity is where I S is the source current driving the Tx, β is the phase constant, − → r is the distance vector from the center of the electrodes to the point where the field is being measured, and [ − → u ϕ ] is the unit vector in the ϕ-direction. The E-field is found by applying Ampere's law where (σ) is the soil conductivity and ω is the angular frequency.
Solving (11) for − → E yields the E-field intensity as a function of the radial distance r The field distribution of (12) has the following three main operational regions [25]: 1) the static region proportional to 1/r 3 ; 2) the induction region proportional to 1/r 2 ; 3) the far region (or radiation region) proportional to 1/r. Both the static and induction regions are related to the near-field of the electric dipole.
In practice, it is the voltage produced within the soil that can be directly measured. The potential difference (Φ) is given by The integral of the dot product produces The exponential integral Ei(−jβr) can be expanded using integration by parts Since r 3 was the highest order term in (12), it is expected that the contributing potential should have a maximum order of r 2 due to integration; taking note that the integration reduces the order such that the radiation region of the potential becomes a constant, the induction region becomes 1/r, and the static region becomes 1/r 2 . All higher-order terms beyond the second order can be neglected. Substituting (15) into (14) and rearranging gives In (16), the terms σβ 2 and −ωεβ 2 are related to the radiation region and are negligible at near-field distances. Similarly, e −j(βr) is approximately 1. Equation (16) can be rewritten as In order for the soil to appear like a conductor, the Tx should be driven within the super low frequency (SLF) to very low frequency (VLF) frequency band (30 Hz to 30 kHz). In this frequency range, ωε becomes small, making (17) more a function of soil conductivity The real term in (18) is extraneous since conductivity in the static region (1/r 2 ) is significantly dependent on temporal charge relaxations [26], not spatial distribution. We can therefore neglect the real component of (17), taking only the imaginary Equation (19) allows one to calculate the approximate voltage radially from the Tx. It should be noted that the E-field is polarized in the − → u θ -direction and thus the cos θ term is typically one. Additionally, since the assumption of this derivation is that the soil acts as a good conductor, the E-field at the soil/air interface must be vertical due to continuity, which again assumes an E-field in the − → u θ -direction and a cos θ term equal to one. Taking (19) and dividing it by the source current (I S ) produces the equation for u e , that can be inserted into (5) and (8)

C. Theoretical and Circuit Validation
To validate (19), and with it (20), the Tx was driven at various currents (I S ) from 1 to 9 A and radial measurements were taken using steel stakes as the Rx electrodes and a Tektronix THS3024 battery powered oscilloscope to quickly measure voltages between two points at the soil surface. The potential differences were measured every 2.5 m, within a 50-m radius around the Tx. The data were then normalized to a 1-m spacing to compare its value with the safety limit of 25 V over 1 m [27].
The theoretical response was calculated from (19) using the soil Φ parameters given in Table I. Fig. 4 shows the measured and theoretical traces plotted side by side. The theory matches the experimental data well, with slight variations due to the simplifying assumption that the soil is homogeneous, whereas in reality, soil is a heterogeneous medium. The results of Fig. 4 indicate (19) can be readily used to predict the potential difference over the area.
With u e verified, the accuracy of the circuit model in Fig. 2 was tested. The electrode spacing was fixed at 1 m, and both front and rear electrodes were moved together away from the Tx at 1-m intervals (see Fig. inset for illustrated layout). This is plotted in Fig. 5(a). Using the measured Tx and Rx parameters in Table I, with a drive voltage that produced 6 A in the Tx; the short-circuit Rx current was calculated from (5) and plotted in the figure. The circuit model approximately follows the experimental data. Next, the effects of the Rx electrode spacing were tested. The front electrode was kept stationary at 10 m from the Tx while the rear electrode was moved backwards in 1-m intervals. Both the measured open-circuit voltage and short-circuit current are plotted in Fig. 5(b). The received voltage at the Rx electrodes was found to be the sum of the potential between the electrodes,  (19). Note that the front electrode was placed 10 m from the Tx and the rear electrode was moved in 1-m increments away from the Tx. allowing one to utilize (19) to estimate the value of received voltage based on the Rx electrode separation at any distance away from the Tx. From Fig. 5(b), it can be seen that as the electrodes are moved further apart, the maximum received power increases. However, the nonlinear distribution of the potential decreases with distance (as shown in Fig. 4), putting a limit on the received power as the Rx is made significantly long (aka the power does not continue to raise indefinitely as the Rx electrode separation is made larger). Long Rx geometries are also seldom feasible. For this study the Rx electrodes were kept at 3 m, as this gave the minimum received power to operate the sensors while being at least a third smaller than the closest tested distance from the Tx (10 m). A unique feature of this power transmission technique is that as the injected Tx current was raised, the sensors could either support a larger load, or they could be moved farther away from the Tx. At 9 A, it was possible to power the sensor modules at approximately 20 m from the Tx.

D. Simulation
With the theoretical and circuit model validated, it is important to develop a simulation that matches the proper conditions of the TTS system. For this work, Ansys Maxwell was used. Future work will utilize this model for investigating better electrode designs, whereas the theoretical/circuit model can be used for quick calculations to determine Tx drive magnitudes and maximum distance sensors can be placed based on the Rx electrode spacing. Fig. 6(a) shows the results from the Ansys simulation for a 1-A drive current into the Tx. A 1-km, homogeneous cube was modeled as the Earth using the same values in Table I. The viewpoint in the figure is top down, depicting the E-field at the surface of the cube in the xy-plane and the equipotential regions, exemplified with dashed concentric rings. For the Rx to develop a voltage, the front and rear Rx electrodes must be located on different equipotential lines.
The simulated values of the E-field were extracted from the simulation and integrated with the distance to obtain the surface potential. The simulated values of the surface potential were then plotted next to the theoretical and experimental values for comparison [ Fig. 6(b)]. It can be seen that both the Ansys model and the theoretical derivation are all in close approximation to the experimental measurements taken in the adjacent agriculture field.

III. POWER TRANSFER FOR IOT APPLICATIONS
A small IoT network of commercial agricultural sensors were used to demonstrate the systems ability to wirelessly transfer power without batteries or large electric storage elements. Do note that a 1-mF capacitor [ Fig. 7(b)] was used for dc filtering and did provide energy for the 0.2-W bursts the sensor required when transmitting data. The sensors were all powered simultaneously within a radius tested between 10 to 20 m around the Tx.
The sensor modules are shown in Fig. 7 and include an integrated moisture and temperature sensor. The microcontroller (μ-ctrl) on each module is an ATMega32u4 with a LoRa communications IC. Fig. 7(a) is an annotated photograph of the sensor module identifying the various components. Fig. 7(b) is a schematic that shows the basic rectification electronics used in the module. A full-bridge rectifier, connected to the Rx electrodes, converted the conduction currents created by the Tx into a dc voltage that was then regulated with a buck/boost converter. When the sensors were powered, soil moisture and temperature data were collected and transmitted via conventional space waves at 900 MHz, to a receiving ESP32 at the base station. The ESP32 was only used as a microcontroller, with an attached LoRa module to accept the 900-MHz protocol. The ESP32 would then relay the data through serial to a computer for storage and postprocessing. Each sensor was waterproofed for long-term measurement studies via a box with a rubberized coating. The sensors were calibrated prior to their deployment. Fig. 8(a) is a photograph of the experimental system's layout with deployed sensors. Illustrations of the Tx and horizontal Rx are inset in the photo for additional clarity. The data were collected for 1 min each hour, continuously over a seven-day period. Fig. 8(b) is a plot of the soil data collected from the four sensors that formed the IoT network.

IV. DISCUSSION ON IMPROVEMENTS IN RANGE AND TRANSFER EFFICIENCY
Equations (8) and (20) predict that the efficiency of the TTS system is strongly dependent on the impedance of the Tx and Rx. Impedance data of the Tx have been collected for several years at the Tennessee location, with the first study on the technique being conducted in Alberta Canada in 2015 [23]. Fig. 9 is a plot of the impedance modulus variations over different seasons, weather types, and locations. In the figure, Fair weather is defined as either full or partial sun, the other weather descriptors are self-explanatory. It can be seen that the impedance of the Tx has a low-Z region and a high-Z region. The high-Z corresponds to a cross-over point where the soil begins to function more like a dielectric. The peaks in this region are a parallel resonance that occurs with the Tx wiring and is observed to have very little energy output into the surrounding media. The high-Z region also fluctuates significantly with weather. The low-Z region not only exhibits the lowest Tx impedance, its fluctuation is between 30-50 Ω and is fairly consistent-even between locations that  are 4800-km apart. It is important to note that both the US and Canadian locations were conducted on crop growing land-the soil thus had a high ion content making it better for plant growth, and consequently more electrically conductive. The data show that the system, while dependent on soil parameters, is quite robust in the SLF to VLF frequency band. The data also indicate a tendency for the system impedance to be lower in more conductive soils with higher moisture, ion, or salinity content. This can be seen in Fig. 9, when the soil moisture is higher, the Tx impedance tends to reduce-noting that snow in Tennessee is nearly always proceeded by rain, so soils are generally highly water saturated by the time snow fall occurs. The system is therefore expected to work better in locations of higher ion content (which is nearly all farming environments) and have lessened performance in drier soils.
Improving the efficiency of the system hinges on reducing the Z T x and Z Rx . It was important to investigate the feasibility of altering the impedance values of at least the Tx. Literature indicates that the resistance of a soil-grounded electrical system can be significantly improved by adding multiple grounding rods [28], [29]. With this in mind, an experiment was conducted where the Tx impedance was measured using only the top electrode (well casing) [plotted in Fig. 10(a)] as the red (circle) trace entitled "no gnd rods." The well casing in the inset illustration was also made red to match the data. The blue (triangle) trace was obtained by installing six 3-m long grounding rods within a 3-m radius around the casing such that all six rods and the casing formed the surface electrode. This new addition was made blue in the inset illustration to match the presented data. The simple addition of the grounding rods successfully reduced the Tx impedance as shown in the blue (triangle) trace of the figure.
The power required to operate four sensor modules was 0.8 W (when transmitting data on 900 MHz). The Z Rx was measured at 250 Ω for the 3-m section between the Rx electrodes. Prior to installing the ground rods, it required a peak input power of 500 W at 60 Hz to operate the sensor network at a distance of 10 m from the Tx. Note that this is the total system input power measured from the solar battery bank into the inverter that supplied power to the Tx. After adding the grounding rods, the same power delivery of 0.8 W was achieved using a 250-W peak, a two-fold improvement in efficiency with an addition of only six extra low-cost grounding rods. Experiments are underway to determine how low the Tx impedance can be made before the bottom electrode would need modification. In its present state, the TTS system is ideally suited for low-power IoT applications with relatively infrequent measurements (such as agriculture). In the experiment, the field was energized for 1 min every hour and shutdown for the remaining 59 min, leading to an average power of only one sixtieth of the peak power or 250 /60 = 4 W. The benefits of instrumenting a field and expending tens of watts on average to know the locations of where to water, fertilize, or spray pesticides can make a significant impact on cost savings Plots of the complex impedance with and without grounding rods. At higher frequencies, the inductive component of the Tx is more prevalent, having a peak inductance occuring at 10 kHz. and the environment. Using renewable sources to power the TTS system also reduces the long-term cost of electricity.
The operation of sensor modules at drive frequencies higher than 60 Hz is currently limited by our drive electronics. However, the low-Z frequency spectrum of the Tx can give insight into such operations. Fig. 10(b) shows that as the frequency increases, the Tx begins to function more like an inductor and less like a resistor; with a maximum inductance occurring at 10 kHz. The addition of the grounding rods appeared to only affect the Tx resistance, not the reactance. This is important since an inductive Tx would not dissipate but store energy in its magnetic field every half cycle, allowing for resonant circuits to be utilized for possible range/efficiency improvements.
The theoretical model also predicts improved efficiency at higher drive frequencies. Using (8), the theoretically calculated transfer efficiency for a single sensor module, as a function of distance (r), is plotted in Fig. 11(a). The plot shows the response of reducing Z T x impedances at three different operating frequencies. Here, Z Rx and R L were kept the same as the experimental data (250 Ω and 56 Ω, respectively). While these theoretical plots of efficiency look promising, we are still in the process of validating Fig. 11(a), and will communicate such validations in future work. The full system efficiency for the proof-of-concept system, while presently low, are similar to values reported for MPT and laser-based technologies. Yet unlike MPT/Lasers, the TTS system is not line-of-sight dependent and does not require the use of complex/expensive receivers. Moreover, the efficiency of the TTS system can be improved by lowering the impedance of the Tx and Rx, which is quite achievable via inexpensive grounding rods. An MPT/Laser system is instead limited by optical-electric conversion processes and beam directionality which has constraints both on safety and complexity/cost of the array.
Furthermore, there does not seem to be a limitation on the TTS transfer range according to the derived theory. Taking (19), a drive current of 200 A at 60 Hz is predicted to enable power transfer over 40 ha (100 acres) to sensors with a 3-m Rx electrode separation (11). 40 ha is the approximate average farm size in the United States using statistical data from [30]. Close to the Tx, the step potential safety limit exceeds 25 V between 1 m. A 40-m radius around the Tx would need to be restricted from access during operation. Beyond 40 m, the step potential is low enough to walk freely. It is possible that the system may cause environmental effects/impacts. However, there has been no negative effects yet observed in the flora or fauna during the course of these studies. Some literature has shown that currents through soil can be beneficial to the growth of plants [31]. It was also discovered that telluric currents generated from a substation 0.8 km away can be observed at the Tx [24]. Measured data show the Tx receives approximately 5.6V Peak from the substation. For a voltage of this magnitude to be measured at the Tx, it can be inferred from theoretical models that the magnitude seen at the substation must be quite large. Since this is an unintentional byproduct of a substation with flourishing trees and vegetation, it can be deduced that the effects on the environment are minimal. Moreover, if a substation can produce effects at such a distance unintended-the possibility of a designed TTS system achieving the same or better level of performance is encouraging.

V. CONCLUSION
In this article, a new LR-WPT technique was presented that utilizes conduction currents through soil to transmit power to sensing devices. A proof-of-concept prototype was constructed that successfully demonstrated power transfer to a small agricultural IoT network within a 20-m radius. The energy transfer is not line-of-sight and simultaneously powers all sensors in the area; making it an ideal tool for agricultural IoT networks. A major benefit of the presented system is that each sensor module does not need an individual battery. Instead, batteries can be placed at a central location and the energy distributed radially around that location without wires.
This work has provided theory, simulation, scaling, and transfer parameters that quantify the operation of the TTS system. The power transfer range was found to be dependent on the injected Tx current while the efficiency was found to be dependent on the Tx and Rx bulk impedance. Increasing drive current while developing methods to lower the Tx/Rx impedance has the exciting potential to wirelessly power devices over great distances. This would revolutionize farm management and transform energy security as higher efficiency and power transfer is achieved. Future work will focus on the development of TTS communications in tandem with power, methods to reduce Tx impedance, and investigations into the effects of frequency and waveform shapes on the power transmission.