Solar Energy Harvesting Research Papers (original) (raw)
This paper carries out an experiment-based channel characterization for a wireless sensor network based on the development kit TI eZ430-RF2500-SEH, which operates with solar energy harvesting. The channel model was obtained from... more
This paper carries out an experiment-based channel characterization for a wireless sensor network based on the development kit TI eZ430-RF2500-SEH, which operates with solar energy harvesting. The channel model was obtained from measurements of the received signal strength (RSS) in an indoor environment. This model is tested through a coverage sensor application, and the energy management of the device is also investigated. Keywords-Wireless sensor networks, energy harvesting, channel characterization. I. INTRODUCTION In the Internet of Things (IoT), a myriad of devices, provided of sensors, will interact in an autonomous and smart manner. Thus, wireless sensor networks (WSNs) will be an essential part of the IoT, as this kind of networks can be widely used to monitor and collect data from the environment, thereby enabling different services and applications. WSNs have been investigated in the most diverse scenarios [1], [2]. For instance, in [1], a two slope, log-normal path loss near ground outdoor model is characterized for a WSN at 868 MHz. In [2], the authors proposed a statistical channel model for a suburban environment, where the multipath and shadowing phenomena are predominant. However, the energy resources in WSNs represent a challenging issue due to the current dependency on batteries. Therefore, energy harvesting techniques have proved to be a promising solution for WSNs [3]. In this paper, an experiment-based path loss model for a WSN based on the TI eZ430-RF2500-SEH kit, which operates with solar energy harvesting, is obtained through measurements of RSS in an indoor environment. This model is tested through a coverage sensor application. In addition, the energy management of the WSN is investigated. The rest of this paper is organized as follows: section II presents the theory used; section III presents the procedure to measurements and the model obtained; section IV presents the effectiveness of the model in a coverage sensor application; section V presents the module's power consumption; finally section VI presents the conclusions of this work. II. CHANNEL MODELING In this paper, we consider two large-scale models, which are described next. A. Log-Distance Path loss Model This model considers that the average received signal power shows a logarithmic decrease with the distance between transmitter and receiver, which is expressed as P L [dB] = P L (d 0) + 10n log d d 0 , (1) where, P L (d 0) is the path loss at the reference distance d 0 , n is the path loss exponent, and d is the distance between transmitter and receiver. Therefore, the received power can be obtained as Pr[dBm] = Pt[dBm] − P L [dB], where Pt[dBm] is the transmit power. Then, for a given scenario, the path loss exponent n can be empirically determined from channel measurements by minimizing the square mean error between the predicted and measured path loss. Hence, the error can be determined as E(n) = k i=1 P Lm i − P Lp i 2 , (2) where P Lm is the average measured path loss, P Lp is the predicted path loss at each point obtained as in (1), and k is the total number of points. By substituting (1) into (2) and differentiating the result in order to minimize the error, the path loss exponent is obtained as n = k i=1 P Lm i − P L (d 0) log d i d 0 k i=1 10 log d i d 0 log d i d 0. (3) B. Log-Normal Shadowing While the log-distance model is deterministic, the random effect due to objects near to the receiver can be characterized using the log-normal shadowing model [4]. This model includes a random variable (RV) Xσ to the log-distance model given in (1), which follows a Gaussian distribution of zero mean and standard deviation σ. Therefore, to empirically determine this model, once the value of n is obtained as described in the above section, the log-distance model is used as the expected value, and the measurement points are projected over that value. Then, σ is calculated as σ[dB] = N i=1 (X i − µ) 2 N , (4) where X i are the projections, µ is the mean of the distribution (0 by considering the path loss model as reference), and N is the number of measurements. III. METHODOLOGY To predict the channel propagation model, measurements of RSS at the anchor node (receiver) of the signal transmitted from a mobile node (transmitter) were taken on 5 different points in order to eliminate both temporal and spacial variation of the channel. To eliminate the temporal variation, at each point, 20 measurements were taken in intervals of 30 min. Moreover, to eliminate the spacial variation, measurements were taken in others 10 points separated 10 cm for each original point. Then, the average RSS for each point is calculated by finding the expected value of all measurements. Thus, from (3) and (4), the experimental values for the corresponding model parameters were obtained as n = 1.0028 and σ = 0.3372.